Is there anything we can do to predict mechanical hard drive failures?

c627627 · Jan 6, 2011

Can software be used to scan a mechanical hard drive every xx weeks to indicate it's failing or are the first signs of it when it actually dies or (if we're lucky) when we start hearing audible knocking...

RJARRRPCGP · Jan 6, 2011

c627627 said:
Can software be used to scan a mechanical hard drive every xx weeks to indicate it's failing or are the first signs of it when it actually dies or (if we're lucky) when we start hearing knocking...

I have a way of predicting HDD failures:

HDDs often give vague symptoms before you lose access.

First symptom to watch for:

Stuttering when scanning a hard disk drive.

Do you ever see it appear to pause then recover? (when scanning and it appears to stop at a one sector longer than the others)

MHDD reports some sectors, when doing a scan (with the F4 key) (not with the "cx" command) with a higher latency than about 150 ms.
And you should worry if you see MHDD report sectors with more than about 10 ms on more recent HDDs!

With MHDD, I can catch real early warning signs of a faulty HDD.

OR WITH HD TUNE: (http://www.hdtune.com/)

A stutter when doing an error scan with HD Tune. Back up if you see stuttering with a HD Tune surface scan,(when watching the numbers, you see a pause, even a short pause, when watching the number counting) even when you don't see red!

The above is how I weed out bad batches of HDDs.

c627627 · Jan 6, 2011

Thank you for that useful post. I believe we need to write up a detailed step by step how to. That's what I would like to see happen.

Mr Alpha · Jan 6, 2011

That is what SMART is all about. Keeping track of hard drive performance and error rates to predict hard drive failure.

RJARRRPCGP · Jan 6, 2011

I seen the SMART of bad HDDs fail to acknowledge that there's a major problem when it's real bad!

Mr Alpha · Jan 7, 2011

RJARRRPCGP said:
I seen the SMART of bad HDDs fail to acknowledge that there's a major problem when it's real bad!

True, SMART is hardly perfect. It has both false positives and false negatives. But it is the best generic hard drive failure predictor i know of.

c627627 · Jan 7, 2011

How does SMART indicate there is a problem?

RJARRRPCGP · Jan 7, 2011

c627627 said:
How does SMART indicate there is a problem?

It's supposed to when there's bad sectors and trouble spinning up, etc.

c627627 · Jan 7, 2011

It's supposed to "how" was my question.

RJARRRPCGP · Jan 7, 2011

c627627 said:
It's supposed to "how" was my question.

I dunno exactly. But it appears that it analyzes sectors during reads and writes and seeks. And difficulties in that are supposed to be counted as a problem.

c627627 · Jan 7, 2011

Counted how, how do are you notified by SMART that there is a problem? Does anyone know?

Mr Alpha · Jan 7, 2011

There are a plethora of SMART monitoring tools out there. Some BIOSes also support SMART monitoring and will throw you a SMART waring during POST if it detects anything wrong.

RJARRRPCGP · Jan 7, 2011

c627627 said:
Counted how, how do are you notified by SMART that there is a problem? Does anyone know?

Based on how much of a struggle it has to do XYZ.

c627627 · Jan 7, 2011

So just because your computer pauses for a bit, you're supposed to just know it's the HD? That does not sound like a good way to be notified. If that is the case, I still think there should be a clear step by step on how to test them on monthly bases.

gofaster87 · Jan 7, 2011

I just touch the drive and feel how hot it is. :burn:

Encore2097 · Jan 7, 2011

Failure Trends in a Large Disk Drive Population - Google Labs
TL;DR: SMART is not really a robust indicator of hard drive failures. Nothing really is. What people have mentioned in the thread are good signs, but are not infallible.

edit: Just had a HD go bad, no SMART errors, just a lot of pauses earlier and seek issues, then bad sectors. I have also had other HD's just stop working one day, others start clicking, etc. My advice: backup your data (preferably off-site or some secure location)

Twigglish · Jan 7, 2011

gofaster87 said:
I just touch the drive and feel how hot it is.

Seems like a never fail kinda idea to me :shrug:

Randyman... · Jan 7, 2011

I'd say run Spinrite on the drive a few times a year. This will 1) Weed out any flakey sectors with the drive's internal error correction, 2) Give you real-time SMART data feedback as the drive is "under load" so you can see how "hard" the error correction is working and 3) Let you know if any un-readable sectors were discovered and recovered.

I use SpinRite mainly for drive maintenance (weeds out bad sectors before they fail) - but it will also recover unreadable sectors - it can even recover parts of a sector! If you notice the SMART Seek Retries and stuff are increasing over time then the drive is slowly deteriorating - and should probably be a candidate for replacement before the SMART stats get too low...

Steve Gibson over at GRC is a Magnetic Hard Drive madman along with other security-related projects! He is a true legacy in our field IMNSHO...

RJARRRPCGP · Jan 8, 2011

c627627 said:
Counted how, how do are you notified by SMART that there is a problem? Does anyone know?

There are just values added and beyond a certain amount is considered a problem.

RJARRRPCGP · Jan 8, 2011

c627627 said:
So just because your computer pauses for a bit, you're supposed to just know it's the HD?

It wasn't my computer pausing, it was a little pause I noticed when scanning, in a program.

Is there *anything* we can do to predict mechanical hard drive failures?

c(n*199780) Senior Member

Member

c(n*199780) Senior Member

Senior Member

Member

Senior Member

c(n*199780) Senior Member

Member

c(n*199780) Senior Member

Member

c(n*199780) Senior Member

Senior Member

Member

c(n*199780) Senior Member

Member

Member

Member

Member

Member

Member

Similar threads

Is there anything we can do to predict mechanical hard drive failures?