Hard drive reliability

[Heihachi_73]

I had a 60GB Seagate last 140000 hours before it randomly died (3/4 of the way through backing up no less) - never had any errors or bad sectors, was copying flawlessly, then suddenly dead, a few clicks upon powering up then power down immediately. The 40GB Seagate in the same PC had 160000 hours and was still good the last time I checked two years ago (no point as I backed it up before the 60GB).

Still got a long way to go with this PC - it's a 320GB Hitachi Deathstar HDS721032CLA362 with just 33684 hours.

[eccerr0r]

how many hours before people typically give hard drives a retirement party?
wonder if i should chance used 55k or 72k POH used hard drives...and how much should be paid for them...

[ChaosLegionnaire]

i just use them till they drop. thats the point of the hardware. all hard drives will fail eventually... just a question of not if but when.

[Topcat]

Originally Posted by ChaosLegionnaire

i just use them till they drop. thats the point of the hardware. all hard drives will fail eventually... just a question of not if but when.

I cycle out drives with critical data usually every 5 years....and to further reduce data loss risk; also run them in RAID arrays for redundancy....but you're right, nothing lasts forever....which is why I can't preach enough for people to keep backups on more than one media; stored off-site.

[eccerr0r]

Well RAID it is, alas 5 years is ~ 44k POH. Just wondering what are acceptable as CHEAP replacements when they do fail... just had one get kicked from my array, it still works alas very flaky. Put in another flaky disk in its place, but it seems to be working better at least. All disks have a lot of hours on them now, at least 40k poh.

[ChaosLegionnaire]

it all depends on what type of data u put on them and your risk appetite. if its critical data then even 1 bad sector is unacceptable because that has already caused loss of your valuable data. if its not too important and/or the data is already backed up on another disk, u can get away with some bad sectors. it is highly unlikely that two disks will develop bad sectors at the exact same spot! so its okay, its still possible to reconstitute all the data back again.

if u're still using flaky disks, then make sure whats on them is backed up elsewhere 1:1. then just use them till they drop. so thats squeezing every bit of $ out of the hardware u paid good money for! hehehe!

[eccerr0r]

Yeah the disks including the flaky disk is in a RAID5, and the array is being backed up to another disk, but it's not completely backed up "sorta" due to the size problem. I'm actually using the other flaky disk that dropped to back up the array since it came back clean in badblocks. Problem is that backups are taken every two weeks or so and there can be a gap between having two disks fail in the array and needing to go back 2 weeks. Was lucky this time, RAID kept uptime up - no downtime unlike when I killed the computer when I ran out of RAM and it swapstormed.

I just need to get another disk at some point so that if a disk gets dropped I can rebuild the RAID5 as soon as I can. It unfortunately takes like 6 hours to rebuild the array. Rebuild seemed to work fine last time when that flaky disk was dropped and swapped in that another flaky disk - but currently all disks are up and running and other than the slowdown during rebuild, it went smoothly - hot swap was successful. Next RAID scrub should be next week I believe, I have each RAID member read every once in a while to ensure parity is good.

Yes, trying to milk every cent out of these disks. Data is semi-valuable (irreplaceable other than backup) but the amount of data is not huge, plus the typical replaceable stuff (few rips, etc.)

[Topcat]

As cheap as large drives are these days, it's hard to beat simple raid1 arrays. Most reliable and easiest to recover in the event of a failure.

[eccerr0r]

Drives need to be even cheaper, especially having to waste one whole disk to protect another - it doubles operational costs. Using one disk to protect two or three disks seems a better use of a disk resources, not to mention a properly designed RAID5 will get some of the benefits of RAID0 during reads.

[eccerr0r]

Woah... big fail on my part.

I was trying to swap out the ailing disk by putting a new disk into an empty slot in the array so I could clone to it. Except I ... oops... opened the wrong door.
I noticed I pulled the wrong door 2 seconds after I saw the disk pop out of slot that I did not expect.

"OSHIT"

I closed the door right away and hurried to check to see what Linux RAID did...

Linux did notice the disk coughed due to it being ejected and reinserted, but Linux-md did not kick the disk!

Whew! Disaster averted!

Placed the preowned disk into the empty slot and started the rebuild. Hopefully in a few hours it will think everything's hunky dory again and I can eject the actual failing disk.

And maybe stick in a hot spare.

[Dan81]

I surprisingly have had luck with Seagate drives recently. Half of my running drives, laptop and desktop, are Seagates, and HGST.

I've had troubles with WD Blues and Toshibas mostly - WD Blues either would fail with slow access times, and Toshibas had their heads go funky. Surprisingly, WD Black (only have one that ORIGINALLY came with a HP G62-a35so - that specific one is fine, HP P/N printed on it), even with SMART failure (had a WD3200BEKT with a Dell part number on it), will keep running at their standard speed of 7200rpm.

[eccerr0r]

Dang just realized... or maybe it's a mistake... my old WD Enterprise RE4 disk is actually 5400RPM! Weird, would explain why I'm getting only 100MB/sec from it. In any case it's still choking and getting random bad sectors everywhere, though it seems to still have plenty of spare blocks, however data is getting lost in the mean time. Wonder if it's possible to get it to a steady state where it's not puking more bad blocks, so I can squeeze more life out of it...

[eccerr0r]

Hmm...also notice that it's losing a lot of sectors about a bit past 75% of the disk... I wonder if I should just tell the drive to behave like a 1.5TB disk or partition away the bad spot... and wonder how much longer I can use the drive with it like this...

[TechGeek]

Originally Posted by eccerr0r

Hmm...also notice that it's losing a lot of sectors about a bit past 75% of the disk... I wonder if I should just tell the drive to behave like a 1.5TB disk or partition away the bad spot... and wonder how much longer I can use the drive with it like this...

I'd boot it from the array before it eats something important in a non-recoverable way. It might be degradation of the head stack, the magnetic coating on the platters themselves, or the head preamplifier is getting weak.

[eccerr0r]

oh it's been gone from the array... it's whether I will continue to use it as a separate drive, trying to avoid the bad spots on the disk...

This is pretty crappy of a drive...only 44K POH, I've had others that have been better.

[televizora]

Usually the drives with higher data density die much sooner than the old drives. With certain exceptions, like when manufacturers changed technologies or firmware/mechanics issues. The big exception seems to be WD Blue. Most WD Blue drives I've seen were ones already with issues/failures. I am yet to see other drive, worse than WD Blue. All with bad sectors. But you do know that Blue means "Sad" too, don't you? Even Hitachi Deathstars are much more reliable than WD Blue.
In my field of work, most of our computers are just "terminals''. Hard drive is required only to boot, open a program, which saves information on a server and work locally with word/excel documents before saving them on a file server. So, hard drive deaths are hardly critical issue for me and I use drives until bad sectors appear. But I do regularily monitor the drives on the servers, where RAID 1+0 is used and I do have backup of the information in case of...two drives from my raid failing simultaneously.
The thing that is worrisome about the hardware RAID controllers that if they do die, even if all the drives are fine, you gonna have extremely bad day.
My opinion about Seagate, Toshiba, even Hitachi drives is much more positive. And WD series different from Blue seem to handle themselves better.

[eccerr0r]

Yay? Been running badblocks -w on this disk until it stops puking bad blocks. It hasn't stopped yet.

  5 Reallocated_Sector_Ct   0x0033   169   169   140    Pre-fail  Always       -
       241
  7 Seek_Error_Rate         0x002e   200   200   000    Old_age   Always       -
       0
  9 Power_On_Hours          0x0032   040   040   000    Old_age   Always       -
       44419
196 Reallocated_Event_Count 0x0032   001   001   000    Old_age   Always       -       214
197 Current_Pending_Sector  0x0032   200   200   000    Old_age   Always       -       0

Was initially trying to reuse this disk in my software RAID5 but this is too unreliable now. Now it's been taken out of the array and replaced with another old disk.
I'm hoping I could squeeze more life from this disk and use this as a backup disk if it can somewhat still reliably retrieve information, or at least if it won't fail at the same spot as the other disk being backed up. Then again if one completely fails it's a problem (which would be implicitly a failure at the same spot...or technically, all spots.)

[eccerr0r]

Nope!

197 Current_Pending_Sector  0x0032   200   200   000    Old_age   Always       -       134
198 Offline_Uncorrectable   0x0030   200   200   000    Old_age   Offline      -       15

Hmm... should try quarantining this bad section next...

BTW what's the "standardized" number of sectors for a 1.5TB and 1TB disk? I've noticed most 2TB disks were standardized across models and brands to 3907029168 sectors or something...

[eccerr0r]

Here we go...

Made a 1.4TB partition (supposedly good), a 250GB partition (bad stuff), and the remainder (about 300GB, supposedly good). Merged the first and third partitions with volume management... and now we'll see how this ~1.7TB volume goes...

[ChaosLegionnaire]

Originally Posted by Dan81

Surprisingly, WD Black (only have one that ORIGINALLY came with a HP G62-a35so - that specific one is fine, HP P/N printed on it), even with SMART failure (had a WD3200BEKT with a Dell part number on it), will keep running at their standard speed of 7200rpm.

yea those wd scorpio blacks are the only good laptop hard drive i've seen. dropped and flipped my laptop over while still in operation and the drive just continues chugging along, no issue, no bad sectors created from manhandling and mishandling my laptop. i wonder whats the secret sauce wd put in their high end consumer laptop hard drive line? its almost as durable and shock resistant like an ssd!! whats the point of upgrading to an ssd then if thats the case? lol!

Originally Posted by eccerr0r

Dang just realized... or maybe it's a mistake... my old WD Enterprise RE4 disk is actually 5400RPM!

what does the model number end in? fyys or fyps? fyps is the re4-gp series, the low power 5400 rpm drive line for enterprise applications. unlike the consumer line greens, those actually work decently and reliably as low power drives without focking with your data but nothing lasts forever i suppose. even enterprise grade drives will die. not if but when.

Originally Posted by eccerr0r

BTW what's the "standardized" number of sectors for a 1.5TB and 1TB disk?

see the attached pics. some images off junkbay to help u.

i noticed that the standardized number of sectors all seem to like ending with 168 for some reason. for 500gb hard disks, its 976,773,168 lba sectors. even the ssds seem to follow this convention! i recently bought a cheapo transcend qlc 500gb ssd and even that has the exact same number of sectors as a 500gb wd scorpio black hard disk i had in my laptop. needless to say, when i found out the sector count matched exactly, i migrated the hard disk to the ssd instead!

however, i didnt junk the spinning clunker disk just yet. i put it on a modular bay hard drive caddy to function as a backup drive now to backup the ssd because when ssds fail, all the data gets hosed and becomes irrecoverable, so its still good to use as a backup drive, just in case. being magnetic, it wont care about the backup writes being done on it unlike nand flash. so yea, use em till they drop, folks! maximize your dollar!