[Discuss] Striping is bad
ma...@mohawksoft.com
well, the does and don'ts of technology. As we all know, sometimes you do
things that you think will "be ok." "It won't happen to me." etc. I set up
the system with striping (RAID0) with no redundancy.
Well, I got burned. I have a DLink DNS-321 NAS box. I've had it for almost
two years. When I bought it, I bought two 1TB Hitachi drives. On
valentines day, drive 1 dies a hard death. No bad sectors, no partial data
loss. The drive would not even spin up.
I took the electronics board from the good drive and tried it on the bad
drive. The motor spun up, but the disk arms would not move (no audible
motion like the good drive).
So, I lost some data. Most of it is scattered across 3 or 4 laptops and my
desktop, but some if it is lost forever.
So, two lessons learned: (1) when you know better, listen to yourself. (2)
don't buy Hitachi hard disks.
_______________________________________________
Discuss mailing list
Dis...@blu.org
http://lists.blu.org/mailman/listinfo/discuss
Alex Pennace
> Most of us have been in the business for some time now and know, wuite
> well, the does and don'ts of technology. As we all know, sometimes you do
> things that you think will "be ok." "It won't happen to me." etc. I set up
> the system with striping (RAID0) with no redundancy.
>
> Well, I got burned. I have a DLink DNS-321 NAS box. I've had it for almost
> two years. When I bought it, I bought two 1TB Hitachi drives. On
> valentines day, drive 1 dies a hard death. No bad sectors, no partial data
> loss. The drive would not even spin up.
>
> I took the electronics board from the good drive and tried it on the bad
> drive. The motor spun up, but the disk arms would not move (no audible
> motion like the good drive).
>
> So, I lost some data. Most of it is scattered across 3 or 4 laptops and my
> desktop, but some if it is lost forever.
>
> So, two lessons learned: (1) when you know better, listen to yourself. (2)
> don't buy Hitachi hard disks.
We recently had something similar happen, luckily the logic board swap
was successful.
At the time we created this array, we went with RAID 0 because the
data that was supposed to be stored on there was temporary in nature,
or there was another copy of it somewhere, a fact we made clear to the
users. We also needed 1 TB of storage, and only had 500 GB drives
available for this purpose.
Or so we thought. The users acknowledged being told that there was no
guarantee, nevertheless they were quite interested in the fate of this
data.
So, we recovered the data, and rebuilt the array with two 1 TB drives
in RAID 1. For bonus points, I moved the logic boards back and sent
the dead drive in for replacement.
So, lesson 3: Even if you think you got your butt covered, you don't.
Edward Ned Harvey
> bounces+blu=nedhar...@blu.org] On Behalf Of
>
> So, two lessons learned: (1) when you know better, listen to yourself. (2)
> don't buy Hitachi hard disks.
(I'll say it the nice way)
How you conclude "don't buy hitachi" from this is situation is ... "Beyond
me."
Derek Atkins
On Sat, February 18, 2012 4:54 pm, Edward Ned Harvey wrote:
>> From: discuss-bounces+blu=nedhar...@blu.org [mailto:discuss-
>> bounces+blu=nedhar...@blu.org] On Behalf Of
>>
>> So, two lessons learned: (1) when you know better, listen to yourself.
>> (2)
>> don't buy Hitachi hard disks.
>
> (I'll say it the nice way)
> How you conclude "don't buy hitachi" from this is situation is ...
> "Beyond
> me."
Why be so nice, Ed? Dude, your conclusions are a non-sequitor for your
data. Or at least your second conclusion. I do agree with #1.
There is nothing wrong with Hitachi drives in general. The fact that *a*
drive failed is just a fact of life -- it happens. I've had mobos fail,
power supplies fail, monitors fail, video cards fail. Parts fail.
Your issue is the fact that you used RAID-0 instead of RAID-1 or RAID-10;
that was just you being silly. I've had a RAID-10 array of four (now 6,
as of about 6 months ago) Hitachi drives running for over four years now.
I did lose one drive which I replaced late last year, and I now have one
spare drive laying around for the next failure. All hitachi. 30,000
hours and still going.
-derek
--
Derek Atkins 617-623-3745
de...@ihtfp.com www.ihtfp.com
Computer and Internet Security Consultant
Jack Coats
If you were thinking that RAID-0 would give you anything other than
a larger single file system
(like ANY redundancy), you were wrong.
There are still reasons for using RAID-0, but you apparently did not
have a reason. Sorry you
had to learn about the reasons for various RAID levels.
I have heard people say similar things about using RAID-5, after
experiencing a double disk
failure. Of course they were not monitoring their RAID-5 (either
visually or by software), so again
the technology worked, just not 'the way they thought it would'.
Education is expensive, whether it is in the University of Hard
Knocks or a 'formal education'.
I just hope your tuition didn't kill a career (I have known some where it did).
><> ... Jack
Richard Pieri
>
> Most of us have been in the business for some time now and know, wuite
> well, the does and don'ts of technology. As we all know, sometimes you do
> things that you think will "be ok." "It won't happen to me." etc. I set up
> the system with striping (RAID0) with no redundancy.
And that's your own fault. You can't blame it on Hitachi. A Western Digital or Seagate disk could have failed in that time span and you'd be in exactly the same situation. I've had WD and Seagate disks die after less than a year in operation. I've had all three brands' disks operational in 24x7 servers for 8+ years. This is not data. This is anecdote, nothing more than single isolated data points. It's like the feedback about Hitachi disks on NewEgg's site. There may be hundreds of reported failures but that ignores the tens or hundreds of thousands of disks that don't fail before their warranties expire.
Disk fail. It is a simple fact. Some fail sooner than others; some seem to last forever. The same can be said about all mechanical devices from CD-ROM drives to automobiles. For example: a few months ago I saw a Datsun 280ZX parked at the local Target. Is this single car, one that's been running for 30-odd years, indicative of the reliability of the 280ZX? Not a chance. It's a single data point, and given the number of ZX's I see on the roads today it is definitely an aberration.
Same with disk drives. Your one failure after two years is just a single data point. Just because the vendor puts a 5 year warranty on it does not mean that you can expect it to last 5 years. It means that the manufacturer has an acceptable estimated failure and return rate for that time span. Certainly some of those drives are going to last more than 5 years. Some won't. That's a fact regardless of who makes your disks.
--Rich P.
ma...@mohawksoft.com
>>
>> Most of us have been in the business for some time now and know, wuite
>> well, the does and don'ts of technology. As we all know, sometimes you
>> do
>> things that you think will "be ok." "It won't happen to me." etc. I set
>> up
>> the system with striping (RAID0) with no redundancy.
>
> And that's your own fault. You can't blame it on Hitachi.
If you are saying that a catastrophic failure of a storage device does not
reflect poorly upon the manufacturer, I suggest you rethink your position.
If I were driving home in a Kia and it died with no symptoms, i.e. was
running perfectly with no "check engine" light as well as properly
maintained, would you NOT blame the manufacturer?
I say that I had the drive for less than 2 years, it was lightly used. The
drive died catastrophically. It had no trouble codes and there were no
issues in the log.
Sorry, yes, for my own sake, I should not have used RAID0, but I do take
issue with the idea that a crap product does not reflect poorly on the
company that produced it.
I will assert that Hitachi drives are poor quality, I've had a 2 year 50%
failure rate in my experience. I have had 4 Samsung drives in roughly
equivalent conditions for a little longer and have had zero defects. I've
had one, and only one, western digital disk die on me in two decades, and
that was because it was in a USB enclosure that got knocked to the floor
while it was on.
Yes, RAID0 is not reliable, but I reject the idea that you can dismiss
faulty product.
Mark Komarinski
> If you are saying that a catastrophic failure of a storage device does not
> reflect poorly upon the manufacturer, I suggest you rethink your position.
>
> If I were driving home in a Kia and it died with no symptoms, i.e. was
> running perfectly with no "check engine" light as well as properly
> maintained, would you NOT blame the manufacturer?
ma...@mohawksoft.com
>> If you are saying that a catastrophic failure of a storage device does
>> not
>> reflect poorly upon the manufacturer, I suggest you rethink your
>> position.
>>
>> If I were driving home in a Kia and it died with no symptoms, i.e. was
>> running perfectly with no "check engine" light as well as properly
>> maintained, would you NOT blame the manufacturer?
> If a Kia went for $100, I'd buy two in case one died unexpectedly.
No a Kia wont go for $100, so your analogy is flawed. You must look at the
economics of it. The Hitachi drives are not cheaper than alternatives. The
objective is to get the best drive for the money.
Make a choice with your $129 per item:
A device that has a sampled 50% failure rate over 2 years.
A device that has a sampled 0% failure rate over two years.
Your choice. If you buy a Hitachi drive, let me know. I've seen some
time-shares in craigslist that I bet I can sell you.
Jack Coats
> On 2/18/2012 10:45 PM, ma...@mohawksoft.com wrote:
>>
>> If you are saying that a catastrophic failure of a storage device does not
>> reflect poorly upon the manufacturer, I suggest you rethink your position.
>>
>> If I were driving home in a Kia and it died with no symptoms, i.e. was
>> running perfectly with no "check engine" light as well as properly
>> maintained, would you NOT blame the manufacturer?
>
> If a Kia went for $100, I'd buy two in case one died unexpectedly.
Sounds like 'mirroring' (RAID-1) to me :)
yep, RAID initially stood for 'Redundant Array of Inexpensive Disks'.
So using anything
above RAID-0 gives some level of redundancy. ... I did use striping
with great results
years ago. But we would RAID-5 2 boxes of 7 disks, and stripe over
the RAID-5 arrays.
Back in that day, 8G 5.25" Seagate disks were 'high tech', and we had
a hardware RAID-5 controllers,
and to get enough room for some seismic data, we did stripe these
controllers together.
We monitored the hardware RAID controllers, and swapped out drives at
the first sign of
issues. ... Still we backed up data regularly.
I understand the frustration with apparent 'bad drives', but even back
then getting 'priority'
Enterprise level drives, we ran into some production runs that had
short life issues. When
it happened, we negotiated with the vendors for 'pre-positioned on
site replacements' being
made available, because we counted downtime in many $$/hour.
Life happens. ... Don't beat yourself up. Just learn and go on.
Yea, the drives could be better, they always can. But you never know
when they can go out
on you, so you have to be ready no matter what drive. Assume they
will go out 'way to early'.
Just be prepared.
My latest thing was wanting to go to 'green' drives (lower power, even
at some performance),
but my luck with them has been pretty bad. They have all had failures
of various kinds.
I just keep an eye on where I can get 'reasonable' replacements
quickly, try to keep my
backups updated, and go on. I live in a 'dusty' environment, and it
just eats drives, no matter
what I do.
Ward Vandewege
> I will assert that Hitachi drives are poor quality, I've had a 2 year 50%
> failure rate in my experience. I have had 4 Samsung drives in roughly
> equivalent conditions for a little longer and have had zero defects. I've
> had one, and only one, western digital disk die on me in two decades, and
> that was because it was in a USB enclosure that got knocked to the floor
> while it was on.
Even within a brand, there are spectacular differences between models of
drives - and even between batches of the same model of drive.
Take, for instance, the 65 Western Digital RE4 2TB enterprise drives
(WD2002FYPS) I took in production about 2 years ago. So far I've got a
failure rate approaching 30% - and most of those died after very light use.
I've seen nothing like this with other WD models, and I've used quite a few
over the years.
Six months ago I installed 55 Hitachi 3TB drives (HDS5C3030ALA630). I have
yet to see a single one of them fail. The backblaze guys also have great
things to say about this model:
http://blog.backblaze.com/2011/07/20/petabytes-on-a-budget-v2-0revealing-more-secrets/
Should I now conclude that all WD drives are completely unreliable and that
all Hitachi drives are awesome?
Of course not. I keep track of other people's experience with certain models
of drives, but always take those accounts with a grain of salt - as someone
else said in this thread, it is all anecdotal information.
Most people do not have a sample size that is large enough to draw any
meaningful conclusions from. And even when someone does have a large pool of
installed drives, how would I know that the batch of drives I buy will be of
the same quality as the one they have?
So I select drives mostly based on length of warranty, accept that drive
failures are a fact of life, and RMA drives when necessary. And I avoid using
anything more fancy than RAID 1 - it's too easy to get burned by failing
drives otherwise.
Thanks,
Ward.
--
Pong.be -( "If you think penguins are fat and waddle, you have )-
Virtual hosting -( never been attacked by one running at you in excess of )-
http://pong.be -( 100 MPH." -- Linus )-
GnuPG public key: http://pgp.mit.edu
Richard Pieri
>
> If you are saying that a catastrophic failure of a storage device does not
> reflect poorly upon the manufacturer, I suggest you rethink your position.
Okay. Hmmm.... no. I still say that you're wrong. A single failure is statistically insignificant.
> If I were driving home in a Kia and it died with no symptoms, i.e. was
> running perfectly with no "check engine" light as well as properly
> maintained, would you NOT blame the manufacturer?
I'd blame the car, not the manufacturer. That is a single fault with a single mechanical device. If, say, 35% of all Kias failed the same way at about the same point in their use lives, then I would blame the manufacturer because that is a statistically significant figure. But just one? Insignificant. Talk to me about reliability after you've put fifty thousand through the wringer and come up with twenty thousand dead within 2 years.
> I say that I had the drive for less than 2 years, it was lightly used. The
> drive died catastrophically. It had no trouble codes and there were no
> issues in the log.
I've had Seagate and WD disks fail just like that after less than a year. No warnings, no trouble codes, nothing. Just up and died. Does that mean all Seagate and WD disks are bad? Of course not. But I repeat myself.
I don't care what the manufacturer of the disks I buy is. Not really. I see disks as consumable materials. They're going to wear out, they're going to fail, and they're going to be replaced, same as the belts, hoses, brake pads and tires on a car.
--Rich P.
Tom Metro
> It's like the feedback about Hitachi disks on NewEgg's site. There
> hundreds of thousands of disks that don't fail before their
> warranties expire.
What would be interesting is if NewEgg reported the percentage of drives
that are RMAed. Though even then it would be a limited metric as the RMA
period ends after 30 days (or some similar short period) and the useful
metric is the drive's reliability over its expected lifespan. But this
would at least tell you which drives have DOA problems, infant mortality
problems, and compatibility problems.
I haven't been there in a long time, but I used to use:
http://www.storagereview.com/
for drive reviews and reliability data. They have a reliability survey
feature, but it still will likely have a negative bias due to being
voluntary. If they could somehow fund it, they would need to do a survey
where they randomly select organizations to participate.
Even with the ideal reliability data being available, as someone pointed
out, that doesn't help you if there are variations between batches or
models. (Although with sufficient data, one could likely draw some
predictions based on how consistent a manufacturer has been in the past.
You might have to narrow your criteria to looking at drives using the
same technology and/or produced in the same factories.)
ma...@mohawksoft.com wrote:
> If you are saying that a catastrophic failure of a storage device does
> not reflect poorly upon the manufacturer, I suggest you rethink your
> position.
A perfectly understandable emotional reaction, but not statistically
significant.
Your argument seems to be predicated on the abrupt nature of the
failure. It presumes that a better manufacture could have designed a
mechanism that would gracefully fail. However, if your particular
failure scenario was never seen in Hitachi's accelerated life tests, are
they still to blame? What if it was seen, but only .01% of the time, so
they decided it wasn't worth adding 2% to the product cost to design
around it. Still Hitachi's fault?
Avoiding failure or failing gracefully can sometimes be possible through
more clever design, but sometimes it just costs more. Other times the
problem is out of the OEM's hands, such as the reliability of a
supplier's component having changed. The manufacturer can be blamed if
they squeezed the costs too much, ignored predictable failures, or
inadequately tested the product. Whether Hitachi has done that can only
be proven through statistics.
Still, perfectly understandable if after seeing a few Hitachi drive
failures that you avoid them for your personal use, but I wouldn't base
a large scale purchase on that data.
-Tom
--
Tom Metro
Venture Logic, Newton, MA, USA
"Enterprise solutions through open source."
Professional Profile: http://tmetro.venturelogic.com/
Jerry Feldman
the risk that a single failing drive can bring down the entire
stripe-set. But, striping is a way to expand an existing volume.
Striping with no redundancy is asking for trouble. By redundancy, I mean
either redundancy by using things like RAID10, or even by doing regular
backups.
The bottom line (IMHO) is that striping is not bad per se, but you need
to understand the issues and ask the questions before you implement.
--
Jerry Feldman <g...@blu.org>
Boston Linux and Unix
PGP key id:3BC1EB90
PGP Key fingerprint: 49E2 C52A FC5A A31F 8D66 C0AF 7CEA 30FC 3BC1 EB90
Mark Komarinski
>> On 2/18/2012 10:45 PM, ma...@mohawksoft.com wrote:
>>> If you are saying that a catastrophic failure of a storage device does
>>> not
>>> reflect poorly upon the manufacturer, I suggest you rethink your
>>> position.
>>>
>>> If I were driving home in a Kia and it died with no symptoms, i.e. was
>>> running perfectly with no "check engine" light as well as properly
>>> maintained, would you NOT blame the manufacturer?
>> If a Kia went for $100, I'd buy two in case one died unexpectedly.
> No a Kia wont go for $100, so your analogy is flawed.
maintenance schedule to a hard drive you purchased online for $129.
Even so, catastrophic things happen, and the manufacturer isn't always
to blame when something goes wrong. Parts die unexpectedly for varying
reasons. A car usually be repaired and put back on the road. Hard
drives can't.
Derek Martin
> Make a choice with your $129 per item:
>
> A device that has a sampled 50% failure rate over 2 years.
> A device that has a sampled 0% failure rate over two years.
The sample size is far too small to produce a statistically
significant result, and any conclusion you draw from it is entirely
based on an emotional response, which could only possibly be correct
by accident.
--
Derek D. Martin http://www.pizzashack.org/ GPG Key ID: 0xDFBEAD02
-=-=-=-=-
This message is posted from an invalid address. Replying to it will result in
undeliverable mail due to spam prevention. Sorry for the inconvenience.