Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[sevenlayermuddle]

Last week my server stopped booting, stuck on a BIOS prompt warning that one of the disks had lots of bad sectors, disaster was imminent, and suggesting that I might want to think about replacing it.  That's  now all done, new disk and no harm done, no data lost and even if there was, it was all backed up elsewhere.   

What  makes it more interesting is that when I built the server about six years ago I initially populated it with two extra Maxtor/Seagate disks of the 7200.11 family.   The failing disk was one of these, the other having failed in exactly the same way several years ago, sudden appearance of 1,000s of bad sectors.   

I know that HDDs do fail, and I shouldn't gripe about a six year old disk failing.   But these two identical disks are the only two disk failures I've personally seen in 20 years of home  computing. 

Co-incidence? 

[kitzuser87430]

Co-incidence?

Don't know really, these may have been made to a price point; they were good value/Gb at the time of purchase and were very popular.

I see a lot of disk failures, I currently have 12 sitting in a pile (from customers laptops).

Recently read an interesting article on the register http://www.theregister.co.uk/2015/01/23/seagate_disks_fail_most/, but as mentioned in the article these were/are consumer disks in 24/7 availability servers.

All food for thought.

Ian

[broadstairs]

I believe laptop HDDs are more prone to failure because they get considerably more rough treatment and probably more power cycles than any in a desktop/server are likely to see so laptop HDD failure is expected more I think. I have only had one HDD failure in over 10 years of using my desktop machines and that was replaced under warranty, prior to that when I was working we saw comparatively few HDD failures in our laptops and desktops but then they were not cheap and cheerful consumer ones

Stuart

[HPsauce]

Oddly in my business I very rarely see HDD failures, but I've had two customers turn up with them in the last week, one a desktop the other a laptop, both HP systems running Windows 7.
One (desktop) was just over 2 years old, the laptop about 4.

Luckily neither was a total failure so some (adequate in both cases) data recovery was possible.
The desktop has been "revitalised" with a new HDD, the laptop has been scrapped as it had other hardware problems and no backup/reinstall media. The laptop HDD was also in a much worse state than the desktop one.

[sevenlayermuddle]

I've just realised, it's worse than I thought.  There's another HDD lying on the shelf, on which I had scrawled 'U/S; 600 bad sectors'.  That is a Seagate ES.2, which is supposedly the Enterprise equivalent and contemporary of the 7200.11.  That means the other 7200.11 is unaccounted for, I can only guess must have chucked it out, in which case it will have gone bad too.  That would then make three out of three failures from Seagate, one supposedly Enterprise grade. 

The Reg article is reassuring for me, as I have been gravitating towards HGST (Hitachi) of late, and these are what have replaced the failing Seagates.  Most recent is a '1TB Ultrastar', purchased from Amazon circa £40 too replace last week's failure.   Only disappointment was the manufacturing date printed on the label, which reads 2010 for a disk sold in 2015.  But at £40 for a 1TB Enterprise (if you believe it) I still think it's a bargain, and probably wouldn't bother to claim under warranty even if I could.

[loonylion]

Recently read an interesting article on the register http://www.theregister.co.uk/2015/01/23/seagate_disks_fail_most/, but as mentioned in the article these were/are consumer disks in 24/7 availability servers.

All food for thought.

Ian

That comment about consumer disks in a 24/7 environment isn't actually relevant, because ALL the disks in the study were consumer disks. You'd expect consumer disks to have higher failure rates than enterprise grade disks, of course, but there were no enterprise disks in that study. Therefore, the environment is irrelevant because all the disks are subject to the same conditions, and the study shows that Seagate are significantly less reliable than other manufacturers. Their bleating about the environment under which the test was conducted is simply trying to distract people from their woeful reliability compared to their competitors.

Seagate were good before they bought out Maxtor (which were terrible at that point), and then Seagate went down the drain.

[guest]

Google (who have some experience in this) said that there was basically no difference in failure rates between consumer & enterprise-grade disks & IIRC they stopped using enterprise-grade disks years ago as the manufacturers could offer no evidence whatsoever of claimed MTBF rates.

I think the same study of AFR (annualised failure rates) suggested that consumers should consider replacing 3 year old+ drives or make sure those drives are properly backed up.

From a personal viewpoint I find that disks which are continually powered up & stay between 25-35C start failing around the 5-7 year mark. Oh and "SMART" is complete rubbish for predicting failures unless the drive got dropped at some point prior to installation.

[HPsauce]

and "SMART" is complete rubbish for predicting failures

Totally agree, there is NO correlation either way in my experience between failures and SMART warnings.

I had a disk the other day that as well as some obvious disk failing symptoms was making the usual "I'm giving up" clicking noises (pretty loud ones too) and SMART said it was fine. I even ran the PC manufacturers diagnostics which have a  specific SMART check in them - all clear!

[Chrysalis]

I have seen many hdd failures in datacentres. At home tho only about 2 or 3.  2 came close together.

In my 2 home failures there was no SMART indicators prior to failure, but in the DC drives there was.  Some of the DC failures were not total failures but me making the DC swap the drives because I was seeing reallocated sectors on SMART.

[sevenlayermuddle]

I'm assuming that those critical of Smart as a predictor, would still agree that sudden and exponential increase in actual bad sectors, that just happened to be made visible by the Smart toolset, would still be taken seriously?

[HPsauce]

that just happened to be made visible by the Smart toolset

If it had been.......
But in my (somewhat limited) experience it hasn't. 

[sevenlayermuddle]

that just happened to be made visible by the Smart toolset

If it had been.......
But in my (somewhat limited) experience it hasn't. 

I've been lucky then.   One was notified by some Linux daemon, by a mail to root or something, don't quite recall.  Last week it was the BIOS that warned me, but in each case it was the sheer volume of reallocated sectors that lent credibility to the warnings.

As a precaution I have also been trying to get SMART self-tests to run on the new disk I just installed, as well as the others.   Seems very temperamental and unconvincing, a waste of time in fact.  But at least 'smartctl'  conforms unambiguously that none of my discs currently have any bad sectors at all, which comforts me.

[Jaggies]

I've just had a Hitachi 1TB drive fail on me - power up-time was 2 years, 1 month according to the disk management tool on Knoppix. Now replaced by a WD 2TB drive, and files restored thanks to nightly back-ups.

In checking the drives with Knoppix, I realised the OS drive has been powered up longer than the Hitachi (I think it's a Seagate) and is showing warning signs of old age, so I've ordered a SSD which should be with me from Wednesday.

I'm hoping for a big performance boost...

[sevenlayermuddle]

One parameter that seems to be missing from SMART is any way to distinguish between 'power up time' and 'time spinning.'.

My server is on 24/7, but I have gone to some trouble to ensure the discs, whilst still being powered, spin-down after a few hours' idle.   That must make a difference, but SMART seems to ignore it.

[Chrysalis]

spin down on idle can actually make things worse as the power up/down process is the most stressful event a hdd has.