Do you test your new hard drives?

[oshunluvr]

I just bought a new 6TB hard drive to replace two 1TB drives holding important data that are now past 6.5 years power-on-time. My experience with drives is once they get past 3 months - they live for 5 years or so, then go. These two are not showing any problems yet, but as I said: Important data and nearly 7 years of use. On this same server are also two 2TB drives that are approaching 4 years power_on_time. Total space in use is about 3.4TB but it's growing at the rate of several 100GB a month.

My plan is to transfer all the data and workload to the new single drive leaving the old drives as backups until they do actually die.

I decided I should put the new drive through some paces prior to expending the energy moving all that data.

I'm curious as to if or how you folks test your hard drives. I've not really done this in the past, but it seems to be a sensible thing to do.

I did discover a nifty little hard drive tester: whdd which I installed from git (after a few needed dependencies) and it's running a read test at the moment (pic attached).

Next I'll run a long s.m.a.r.t. test. Tips? Methods? Comments?

[ronw]

I own a copy of Spinrite, and run it on new drives. Level 2 (read only) on SSDs. I'm not sure it is worth the price if all you run are SSDs. It has pulled a few of my spinning rust boxes from the jaws of death over the years.

[SteveRiley]

"ETA 528:38." Nine hours -- that's a long time.

TBH, it isn't something I've ever thought about for factory fresh drives.

[Rod J]

I always install GSmartControl (gui for smartmontools) soon after I install an OS. I've found it invaluable for testing new drives and old drives which are showing problems. I have SMART monitoring turned on in the BIOS too of course. Generally, if the SMART attributes are deteriorating (shown as pink in GSmartControl) it's time to think about replacing the drive or using it in a non-critical way until it dies.

One caveat though, It seems to me that a drive can still pass a SMART Health Check status but really be just about dead. I've generally replaced the drive long before it ever fails the SMART Health Check.

Oshunluvr: I like the look of that app "whdd". It looks like it will give a lot of detailed info on a drives' condition.

[Qqmike]

So far, I have found that if the hard drive is not basically DOA (i.e., it makes it past 1-3 months), and if it goes 1 year, then it goes indefinitely! Nine+ years so far. Seagates, WD's. Good practice, though, would be to replace every 4-6 years, imo, to be safe, especially if it is mission-critical somehow (to home or business). Somewhere around here, dibl had posted a nice little spiel on SmartTools (or whatever that is called), I tried it once, works great.

[Qqmike]

How to Test Your Hard Disk Drive

https://www.kubuntuforums.net/showthread.php?53995-How-to-Test-Your-Hard-Disk-Drive

[dibl]

I'm too lazy to invest the time to test a new drive, but I agree it is a good idea. Except for a small (and cheap) SSD that was used only for booting, it has been decades since I experienced a drive failure. But I don't mess around when I get suspicious about a drive -- my data are too precious to me to take any risk at all. I ditched a pair of WD1000's a couple years back when the BTRFS filesystem that ran on them coughed up a few errors. They only had about 3 years' of running time, and I would suppose only one of them was having any problem, but I'm not interested in taking chances. The replacement pair is humming along nicely with no errors as of today.

[InsideJob]

I like gnome-disk-utility (just Disks if you actually use gnome instead of KDE). Not only does it do all the SMART control stuff but it can turn off my external hard drive before I unplug it. All "safe remove" in KDE does is unmount. You can easily drop a spinning drive on the floor fumbling with USB cables. Doesn't really matter how it tested before that. Doh!

[SteveRiley]

turn off my external hard drive before I unplug it

No need to install GNOMEish stuff just for that. The following command, where X represents the device in question, accomplishes the same thing:

sudo hdparm -Y /dev/sd[i]X[/i]

[InsideJob]

I prefer gparted to KDE's partition manager too. Did I mention the gnome-disk-utility also does images? So it's like dd, smartcontrol and hdparm all in one... and you don't have to read man pages to figure out cryptic switches.

[vinnywright]

I prefer gparted to KDE's partition manager too. Did I mention the gnome-disk-utility also does images? So it's like dd, smartcontrol and hdparm all in one... and you don't have to read man pages to figure out cryptic switches.

KDE's partition manager will do partition "back up's" to an image file or a copy ,,,,,,,

VINNY

[oshunluvr]

I too, like the gnome-disk-utility. I wish it was ported to QT, however it pulls in very little gnome-ish stuff.

Depends:

3.14.0-0ubuntu2 - libc6 (2 2.15) 
libcairo2 (2 1.2.4) 
libcanberra-gtk3-0 (2 0.25) 
libgdk-pixbuf2.0-0 (2 2.22.0) 
libglib2.0-0 (2 2.37.3) 
libgtk-3-0 (2 3.9.12) 
liblzma5 (2 5.1.1alpha+20120614) 
libnotify4 (2 0.7.0) 
libpango-1.0-0 (2 1.18.0) 
libpangocairo-1.0-0 (2 1.14.0) 
libpwquality1 (2 1.1.0) 
libsecret-1-0 (2 0.7) 
libsystemd0 (0 (null)) 
libudisks2-0 (2 2.1.1) 
dconf-gsettings-backend (16 (null)) 
gsettings-backend (0 (null)) 
udisks2 (2 2.1.1) 
adwaita-icon-theme (0 (null))

Looks like most of that is not GTK.

Anyway; The final results of the whdd read test were no bad sectors and only a few really slow ones. I'm calling it a success. Running a long smart test now. It will take even longer than the whdd read test did - 707 minutes!

Re. drive history: Back in the old days of half-height and full-height 5.25" drives - remember those monsters? - I had an old full-height drive that I pulled from a discarded PC that I parted out. It had to be pushing 10 or 11 years old. I don't remember the capacity. I finally just threw it away even though it was still working! More recently, I built a new desktop about 9 years ago. It had four 500GB WD "Blue" drives and I ran a speedy RAID on those for years. Eventually relegated as backup drives and I re-purposed one of the four into a PVR, they were still chugging along until last month when one of the three still in my desktop started throwing smart warnings. I had already lost one to an accident, so I took the errors as as sign and dumped the last one as well. They weren't really being used for anything anymore anyway. SSDs and much larger capacity (and cooler and more energy efficient) drives had already taken their duties.

[Rod J]

I too, like the gnome-disk-utility. I wish it was ported to QT, however it pulls in very little gnome-ish stuff.

Yeah, I remember using gnome-disk-utility back when I was using Ubuntu Lucid. It is useful, I too wish it was ported to QT. I decided to install it just now and it pulled in minimal stuff (I probably had installed other things that had gnome dependencies already).

Re. drive history: Back in the old days of half-height and full-height 5.25" drives - remember those monsters?

Showing my age here now! My first hard drive PC (about 1989) had a whopping 10Mb hdd! Yes, young ones, not 10Gb but 10Mb! Back in the DOS days it wasn't so bad though. Later on, I remember someone going on about having a 1/2 Gig drive ... I was truly impressed . Back then a 500Mb drive was truly huge, about 4 - 5 inches thick with 5 or 6 platters I think and made so much noise it was ridiculous thinking about it now. Ahh, the good (bad) old days!

[oshunluvr]

Ok, drive passed both tests. I guess I'm putting it in action!

[InsideJob]

I remember having to manually park the heads of my ESDI hard drives (they were cheaper than SCSI.) Heck, I remember seeing my first real Winchester hard drive plugging into a Radio Shack TRS-80 Model II back when they had Computer Centers in addition to the regular stores. They'll probably go the way of CompUSA and Circuit City soon.

The mechanical parts are tested by the manufacturer and these days the technology is really reliable, with auto-parking g-shock and Teflon encase heads and such... It's the controller board screwed to the bottom that's most likely to fail on a new drive. I read somewhere that if a capacitor or resistor is going to fail it'll usually do it in the first 24-48 hours of operation. So a couple days of "burn-in" before using in production seems sufficient, IMVHO.