Tired of failing hard drives

English Hardware
1

I am about to replace my third western digital drive inside 6 months.

The drives are all SATA3 1 or 2 terabyte drives in a home server that runs 24-7. I’m used to getting up to 7-8 years from a drive but these have all failed within a few months and I’m sick and tired of rebuilding my server.

It’s been running Open Suse 12.1 with KDE3.

Each drive has a cooling assembly with two fans, and they’ve been running at about 31-32 deg C, which seems to me is quite cool for an HD.

The last three drives I have replaced have all been Western Digital Caviar Green 1 or 2 TB drives. WD has been quite good about replacing them, but I don’t have the time to keep doing this.

Does anyone have recommendations or a url for large scale tests of hds?

2

Hi
Don’t use green drives for something that’s running all the time… I
use WD RE3’s or better…
http://paste.opensuse.org/7632794

All mine run at around 30C

Install iozone for benchmark testing.

–
Cheers Malcolm °¿° (Linux Counter #276890)
openSUSE 12.2 (x86_64) Kernel 3.4.6-2.10-desktop
up 1:36, 4 users, load average: 0.07, 0.21, 0.40
CPU Intel i5 CPU M520@2.40GHz | Intel Arrandale GPU

3

So, unless we are talking about SSD drives, who’s left that makes physical hard drives? Western Digital & Seagate is all I can think of. Maxtor went to Seagate, Seagate and Samsung are aligned and Western Digital got Hitachi and they all eat someone else way back when. Next, hard drives do fail and its not if but when they will fail. Heat and vibration are the hard drive killers, but hard drives just fail on their own at the day and time of their choosing which is why you either need to backup often and/or keep lots of separate copies of everything that is important to your digital life. In the last year, my last hard drive (not SSD) to fail was a WD and the last hard drive problem was a Hitachi external unit where the USB to SATA interface died even though the hard drive was still working. And I even had a SSD drive pass on to the next life sold with the name of Corsair, though I am not sure who really makes it. So I could to say you are very unlucky or, the cooling you have is not working as you had hoped or its just another normal day for anyone that owns a hard drive. For any computer having lots of problems, consider it might be time to invest in a better power supply. And of course, it might be time to switch over to Seagate while you still have that choice to make. None of this I say is reassuring, but really, no one knows for sure why a hard drive just died that was only six months old, more than once in the same PC. But, as you can see, it does happen to the best of us. The more hard drives you have, the more likely one is going to die on you today.

Thank You,

4

On Mon, 10 Sep 2012 00:56:01 +0000, robertsmits wrote:

> Does anyone have recommendations or a url for large scale tests of hds?

For stuff that runs 24x7, I wouldn’t look at consumer hard drives, but
rather at hard drives used in data centers and in gear intended for that
use.

Gear designed for high availability use is going to generally last
longer, but that additional lifespan comes at a higher price.

Jim

Jim Henderson
openSUSE Forums Administrator
Forum Use Terms & Conditions at http://tinyurl.com/openSUSE-T-C

5

On 09/09/2012 10:00 PM, Jim Henderson wrote:
> On Mon, 10 Sep 2012 00:56:01 +0000, robertsmits wrote:
>
>> Does anyone have recommendations or a url for large scale tests of hds?
>
> For stuff that runs 24x7, I wouldn’t look at consumer hard drives, but
> rather at hard drives used in data centers and in gear intended for that
> use.
>
> Gear designed for high availability use is going to generally last
> longer, but that additional lifespan comes at a higher price.

Have you checked the output of smartctl to see if you are exceeding some
particular parameter. I had a WD green 1 TB drive that was unloading the heads
too often. By resetting a parameter, I cut the occurrence way down.

6

Hi
For sure, these aren’t too bad;
http://www.newegg.com/Product/Product.aspx?Item=N82E16822136798

I’ve got two 36GB 10K raptors brought in 2003 that are still
working…

–
Cheers Malcolm °¿° (Linux Counter #276890)
openSUSE 12.2 (x86_64) Kernel 3.4.6-2.10-desktop
up 2:55, 4 users, load average: 0.24, 0.10, 0.06
CPU Intel i5 CPU M520@2.40GHz | Intel Arrandale GPU

7

Hi
For sure you can use hdparm to check and set;


hdparm -B /dev/sdX
more detail
hdparm -I /dev/sdX

From memory setting the APM_level to 254 disables.

–
Cheers Malcolm °¿° (Linux Counter #276890)
openSUSE 12.2 (x86_64) Kernel 3.4.6-2.10-desktop
up 3:17, 4 users, load average: 0.24, 0.10, 0.06
CPU Intel i5 CPU M520@2.40GHz | Intel Arrandale GPU

8

> Does anyone have recommendations or a url for large scale tests of hds?

i don’t have a url, but a person i know to be the administrator of
several servers for many different companies in his local area wrote to
another forum just last week (about 24x7 drives)

the universe of drives that I’m willing to buy is the Barracuda
ES, Western Digital Caviar Black, Hitachi DeskStar if it has a 5-year
warranty, and Samsung Spinrite if it has a 5- or 7-year warranty.
But if I’m actually planning a mirror from scratch, I always get one
drive from each of two of those families.

that is he only buys those drives today (last year he had a different
list) and he no longer uses a matched pair in RAID since too often
both/all fail at about the same time!

and a couple of years ago he found that you have to carefully read the
manufacturers spec sheet because it is not always easy to find if you
are buying a drive made for use as 24/7 or for an work/rest cycle…and,
it is NOT good to use a 24/7 drive in a machine shut down each night, or
an work/rest drive in a 24/7 server!!! either case will result in
shortened service life!

–
dd http://goo.gl/PUjnL