Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

After a hard drive fail (lots of corrupted sectors) I reformat it, but the drive give me no confidence to use it in production.

  1. Is there a tool to stress test so I know I can use it?
  2. In which ways would you use the drive so if it fails again you don't end up with data loss?
share|improve this question

13 Answers 13

up vote 20 down vote accepted

The problem is that if a drive has failed you once, you can't trust it. In almost every case, the value of the data you are proposing to store, plus the potential costs of loss or recovery if that data gets lost, exceeds the cost of a new drive.

If the data you are proposing to store has any value to you at all, don't use the drive. Buy a replacement.

I might consider using it in a RAID-1 if I was desperate, and even then probably only for 100% scratch (ie local working storage on a compute node).

The alternative is to convince yourself you have a controller issue, and that's why the disk got corrupted. If you buy another disk and almost immediately have the same problems, then that's possible, and you can probably use the first disk in another computer.

share|improve this answer
6  
+1. Disks are cheap. Once it's failed once I wouldn't trust it with any data I didn't mind losing. –  theotherreceive Jul 13 '09 at 14:14
    
+1 Disks are WAY too cheap to be bothered with a failed drive. Just to clarify though, some people think a "failed drive" is when some file goes missing, gets corrupted, or some other "voodoo". I know of at least one situation where a business was going to replace an entire PC (and HDDs) because they had a virus that wiped out some Windows system files. Best Buy told them it was a "bad hard drive". ;-) –  KPWINC Jul 13 '09 at 16:00
3  
This is why it's important to run a low-level diagnostic against the drive to make sure there isn't just some fluke corruption. Simply tossing the disk without diagnosing the true problem doesn't make a lot of sense. –  Russ Warren Jul 13 '09 at 16:06
2  
When Best Buy is a business's tech support, that tells you a lot already. –  David Jul 13 '09 at 16:35
    
seriously, who the heck is desperate enough to use a failed hard drive... ever? –  Keith Jul 17 '09 at 19:02

I like to run SpinRite on semi-defunct drives just to see if they're worth anything. It takes quite a while, but it's worth it in the long run.

See here: http://www.grc.com/intro.htm

share|improve this answer

RMA it under warranty!

I can't believe no one has said that yet.

It probably has a 3 or 5 year warranty. Send it back and get a new one (or refurb). Don't be afraid of refurbs. Many drives get returned to the mfg with zero trouble (BIOS problems, software scrambled the data, bad controller, funky software RAID, etc).

Whatever drive you have, new or refurb, check for bad sectors.

share|improve this answer
2  
Just as long as you know there's nothing particularly sensitive on the drive. –  Brad Ackerman Jul 13 '09 at 19:38
1  
Good point, Brad! DBAN comes in handy here, since we are assuming the drive is still somewhat usable (according to the question). –  kmarsh Jul 13 '09 at 19:53

If a drive has lost data, the best way to get some use of it is take it apart, and extract the magnets. They make great fridge magnets.

Hard drives are cheap, it's not worth taking a chance.

share|improve this answer
4  
+1 for the magnet comment. They are SERIOUSLY STRONG magnets and a lot of fun. Be careful though, they're extremely brittle as well. –  KPWINC Jul 13 '09 at 15:55
1  
If you don't remove them from the metal plate, they're strong enough and shaped right to hold up things like oven mitts. On the oven, not the fridge, but the idea is the same. –  chris Jul 13 '09 at 21:55

In which ways would you use the drive

  • Paperweight
  • Doorstop
  • Pull it apart and have fun with the magnets
share|improve this answer

Better than just reformatting would be to overwrite the whole disk with data. Then the drive has a chance to remap flaky sectors with spare ones.

You could check the reallocation count in the SMART data (e.g. with smartmontools) before and after overwriting to get an idea how many sectors were bad. You should also check regularly to see if the reallocation count went up.

In theory formatting the disk should be enough to make sure that new data is written to disk but I wouldn't trust that some tool wouldn't try to read an "unused" sector and tripped over a read error. Also its nice to know the amount of bad sectors beforehand.

If you don't trust the disk any more use it in some kind of RAID setup i.e. in a mirror or RAID5. Not trusting a single disk is always a good idea ;-)

share|improve this answer
    
+1 for mentioning smartmontools. –  kmarsh Jul 13 '09 at 17:39
    
At the very least, it would ensure that no sensitive data was on the drive for someone else to recover, after you had thrown it out. –  Ernie Jul 13 '09 at 19:15

Never trust a failed disk in a RAID array. Ever.

I had this happen to me plenty of times, I didn't bothering setting up Nagios to monitor the hard drives SMART status and hard drives kept reporting failures (plead guilty for not doing any or not knowing actually) but nothing would happen to the server, it would remain up and working.

Until the hard drive really starts pucking up. The whole RAID array starts getting corrupted instead, happened with 3Ware and Adaptec controllers, then I end up having to fsck the file system with a lot of corrupted files, which seemed like an eternity when the clients are begging for the server to be back.

So I really would recommend just using it somewhere that you don't care about your data, or if the hard drive is not that old, RMA it. Both Seagate and WD have impressive RMA procedures and you usually get a refurbished hard drive, sometimes with even bigger capacity.

However, don't touch Seagate hard drives. I have lost all my trust in them, I have had a good 25% failure rate of those hard drives, from different distributors too. NS or AS version, they all failed. I have been using WD RE3 and it's been great.

share|improve this answer

Yah, disks are so cheap now. It all depends on how critical the data stored on it is. I might just use as a secondary backup or just a scrap drive. Personaly, I would not risk any data stored primarily on it. I have had too many disks crash to ever care about unreliable drives again.

Once I know a drive gives me issues, the sledge hammer of justice delivers!

share|improve this answer

If you can boot from a CD or USB stick, I suggest you try GRC's SprinrRite

not free, but IMHO well worth the price!

share|improve this answer

If you use *nix, do a dd if=/dev/urandom of=/dev/disk on the disk a few times and see if SMART or syslog gives anything.

Disks are so cheap these days, just go and get yourself a new one. :)

share|improve this answer
    
This doesn't check the drive (directly) but destroys it through overwrites. You could test a read cycle (using "dd if=/dev/disk of=/dev/null") and not destroy the data. –  Mei Jan 13 '11 at 22:09

You may want to look into SpinRite, which does many different things related to hard drives.

You could also grab a drive fitness test disk from the drive manufacturer to see if any errors show up. The Seagate and Maxtor tools are here, for example.

As for using the drive, you could use it as scratch space where nothing is stored permanently, or just use it on a system where you know everything is backed up somewhere else.

share|improve this answer

If a disk even looks like it might fail, you need to toss it and get a new one ASAP. HDD's are probably one of the top two components most likely to fail in a running system (the other being cheap crappy power supplies in my experience). Unless the data on the disk is absolutely worthless, you need to replace it with a new disk. (eg you are using it as a temporary cache/proxy, and you don't care if you lose the data or not)

share|improve this answer
    
I can't think of any situation where corrupted data would be allowable (swap? corrupted binaries... cache? corrupted entries...). You don't "lose" data (unless the metadata goes bad) - you get corrupted data. –  Mei Jan 13 '11 at 22:10

The value of the data is worth a new drive. Unless you absolutely have no budget for a new drive, id go and get a new one.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.