Testing External USB Drive Prior To Extraction

In a moment of weakness I bought a 5T Seagate Expansion drive with the plan to open the external case and harvest the internal drive within. I typically don't buy externals because doing this voids the warranty. But the 5T is not yet available as an internal, and decided I'd figure out some way to test the drive prior to extraction while the warranty was active.

 

So now is the time.

 

I am interested in any suggestions. One thing I feel is very important is managing disk temperature. Others may disagree, but I feel that running a drive hard with high temps is going to negate any benefit of this exercise.

 

So here are a few strategies I'm considering, presented in no particular order ...

 

1 - use preclear with a fan blowing directly on the drive in a cool environment. That might keep the temps cool enough. Problem is preclearing 5T using the USB2 drivers in unRaid would be painfully slow. Maybe someone could suggest an easy to boot Linux with USB3 drivers that I could use.

 

2 - similar to #1 but making some enhancement to preclear to go into a waiting mode if the disk temp crossed some threshold, waiting for it to go under the threshold by a few degrees before resuming

 

3 - abandon the idea of testing every sector and instead run an abbreviated test of say 10% if the surface. Preclear could test every 10th block and in essence only test 500G instead of 5T.

 

4 - write a custom program that can run under Windows to simulate the three primary things preclear does - reading every sector, writing every sector with zeroes, and reading/verifying every sector. It could be smart enough to run the smart reports and do the pausing thing on high temps, if needed

 

5 - stop being so anal and open the external drive. Chances are very low that your external is bad. Maybe run a long smart test and then be done with it.

 

6 - locate some other tool that would be good for this purpose. A quick scan of tools did not identify anything I was confident would work on drives over 2T. And nothing had a feature to pause the test on high temps. But maybe someone here has some knowledge or experience with a useful tool.

 

7 - before unRaid I used to run a test of sorts on new disks. I'd create a 100 byte file of "x"es with copy con. Call it "a". Then do "copy a+a+a+... b" (10 a's). Now b is 1000 bytes. Repeat copying 10 b's to c. Etc etc. Doesn't take many times to be up to very large files that can be used to fill the disk. This would not have the ability to regulate temps, but could be run on a Windows box manually checking temps (with a cool blustery environment temps may not be an issue).

 

I'm interested in what others do, have done, and any Rube Goldberg machine ideas from armchair quarterbacks to give my USB drive a good workout prior to freeing the internal and voiding my warranty in a reasonable time frame. Leaving it in a precleared state is not important as a new parity build is required.

 

Thanks!

As I replied in another thread I did the following.

 

I installed VirtualBox on my Windows 7 PC with USB3.0 ports.  I setup an unRAID VM in VirtualBox.  I RDM'd the USB3.0 drive to the VM.  Ran the Preclear script.  It looked like it was going to work but I didn't try to cool the drive and the temps got to mid to upper 50's and the preclear never finished.  I'm blaming the high temps on the failure but it is possible it wouldn't have succeeded even if I had used a fan to keep the drive cooler - don't know.  The speeds I was getting started out at 70MB/s so about half normal speed but still faster than USB2.0 directly from unRAID would have been.

 

I run preclear with a fan on the drives. Yes, it's slow and noisy, but once you shuck a drive and it fails, you'll spend the time. I keep thinking to apply some duct work, but have not yet.

Bob - I use VMware. Not sure what is RDM? UnRaid I think would see the drive as USB and take just as long as if connected to unRaid server. I'm concerned to install vitualbox with VMware.

 

C3 - what ductwork? What temps are you seeing with fan blowing on drives?

Duct work in my case would just a cut up cardboard box to guide the air into and out of the enclosures, nothing special, but I have yet to bother. The temps on my drives are very safe, in the 30s.

Bob - I use VMware. Not sure what is RDM? UnRaid I think would see the drive as USB and take just as long as if connected to unRaid server. I'm concerned to install vitualbox with VMware.

 

C3 - what ductwork? What temps are you seeing with fan blowing on drives?

See if this answers your questions: http://lime-technology.com/forum/index.php?topic=7914.msg76501#msg76501 although that is about ESXi the concept is the same.  Also good is the VirtualBox manual here: https://www.virtualbox.org/manual/ch09.html#rawdisk.  It explains what to use for Windows Hosts as well but here is what I used to create the vmdk RDM image pointer(for lack of a better name):

VBoxManage internalcommands createrawvmdk -filename "v:\RDMs\Preclear.vmdk" -rawdisk "\\.\PhysicalDrive18"

Then I attached the vmdk file to the IDE controller of the unRAID VM.

 

 

But no unRAID would see it as an IDE Hard Drive doesn't know it is USB at all.

Bob - I am going to study that and try to figure out if I could do something similar with vmware.

 

But for this time I went with c3's suggestions / experience.

 

I found a little bug with myMain (not really a bug, but it had issues with the S/N on the drive which had some consequences trying to set the "smartopt" custom drive parameter). I fixed it and will send the update to Joe L. with my next update.

 

I set smartopt to "-A -d sat". This worked for my 5T Seagate Expansion.

 

I ran preclear with the following options ... "-A -d sat -b 1000".  I got some warnings on the smart report. Not sure why smartctl returned a "4" return code, all of the data seems to be there and correct. The -b 1000 means that it reads for a very long time before taking a breather and refreshing the screen, speeding up the preclear a bit.

 

So it has now been running for 22 minutes and has read 50G so far (1%). That's about 37 MB/sec. This will take a while. Temps slowly creeping up from 31C to 34C. I have 2 120mm fans blowing on the bottom of the drive which is on it side. I doubt we'll get near 40C. Probably the slow speed keeps it cooler.

 

I will let it do its thing.

No errors so far.

Yawn. The drive is now 88% (pushing 4.5T) through the pre-read at 33 hours in. I project it will take a total of 4.5 days to complete the single pass, so should be done on Monday evening after work.

 

No errors and drive temps have not exceeded 34C. Constant 37 MB/sec speed.

 

I found a low quality video explaining the extraction process for a 3T Seagate Expansion that looks the same at the 5T. I plan to try to follow them. He uses 4 old credit cards, but even this experienced guy seems to struggle and risk crunching his knuckles in the process, taking him over 10 minutes to do it. I saw another video with a different unit where a guy used guitar piks. What you you found that works best? Any other tips? Is it possible to be careful enough to get it disassembled so it could be put back together for a warranty claim if required? It looks to me like the extraction breaks the clips that hold the USB housing together.

You want to use something softer than the plastic used to make the enclosure. That way the tool gets damage, not the case. I use pieces of plastic cut from the "blister pack" of recent purchases. Blister pack is that heavy clear plastic that makes opening things such a pain. It is a bit softer than guitar picks or starbucks stirrers, but in good supply cheap (I don't drink coffee). I can cut to what shape is needed and make as many as needed. I still have a pile of sheets cut from the Seagate 4TB Backup Plus packaging.

 

The key thing is to realize the tool should be damaged by the process. This means multiple and replacement tools over time. It will take time, but 5 to 15 minutes is typical, longer the first time. And watching the videos is highly recommended so you can learn where the clips are.

 

If you do successfully open it without any damage, and there is no tamper detection tape (started seeing in some), you get to save the enclosure for 5 years (since who remembers to toss it in 1 year).

My preclear in the USB enclosure finished. More than 125 hours!

 

And I used my optimized post read tool, otherwise it would have been 10-15 hours longer.

 

No errors.

 

Would be nice to see USB3 support added to unRAID.

 

Tonight I will try to extract the inner drive.

I successfully freed my 5T drive from its case. Easier than expected. I am not convinced I could reassemble it such that Seagate would honor a warranty, but maybe I'd have a chance.

 

After preclearing on the USB, it is testing properly as precleared.

 

I did a little read test and speed was about 120 MB/sec about 1/2 way through the disk. That is fast enough for my needs.

 

Only one strange thing. The drive did not report its spin state with hdparm -C.  When I ran the command it returned "unknown".

 

Then I ran "hdparm -y" to spindown the drive, and then it reported its state correctly. And then when I woke up the drive it reported its state correctly.

 

Strange. Anyone ever seen something like this before?

 

I read some reports of extracted 5T drives not working (hanging) Windows 8.1 But unRAID seems to like them.

This is an ongoing issue. The drive is not in my array yet, but if it sits for a period of time it will no longer report its spindown status properly.

 

Not sure if this will affect operations in unRaid or not.

 

I also posted a question in the preclear thread. The drive size is 4k smaller when it mounts as a USB disk versus as a SATA disk. Not sure if this could impact unRAID's acceptance of the drive as precleared.

I started a preclear on my 2 Seagate 5TB external drives and ran into this same slow issue connecting to USB 2.0 connection.

 

I didn't see any easy way to take these apart or test with a SATA connection which I was able to do with the Seagate 4TB Backup Plus external drives.

 

This was going to take "forever" at 22.6 MB/s!

 

Here is what I saw at 50% complete:

 

Pre Read in progress on /dev/sda: 50% complete.

(  2,501,482,905,600  of  5,000,981,073,920  bytes read )at 22.6 MB/s

 

Using Block size of  8,388,608  Bytes

Next report at 75%

Calculated Read Speed: 22 MB/s

Elapsed Time of current cycle: 30:48:49

Total Elapsed time: 30:48:50

 

I read that unRAID 6.0 beta supported USB 3.0, but my MB does not. So I went to Micro Center and picked up an Inland 2 port USB 3.0 PCI Express card for $9.99 and updated my test system to 6.0beta12.

 

That worked and now I'm getting over 150 MB/s while connected to USB:

 

Pre Read in progress on /dev/sdb: 50% complete.

(  2,500,771,840,000  of  5,000,981,073,920  bytes read )at 155 MB/s

Disk Temperature: 35C,

Using Block size of  1,003,520  Bytes

Next report at 75%

Calculated Read Speed: 156 MB/s

Elapsed Time of current cycle: 4:26:16

Total Elapsed time: 4:26:17

 

Wasn't sure that it would work with USB drivers needed for the Inland card but it did and really sped up the Preclear!