Hey ALL,

 

I had two drives fail this year, a 500Gb WD 5000KS and a 4000KD. I did a 750Gb swap for the 400 with WD in which I just paid the upgrade fee and they pre-replaced the 400 for me and I did not have to send it back but they wiped the serial number from there system so it no longer has warranty. The 500Gb drive I was ready to do the same with but I thought I would finally get around to testing the drives. I setup a test-bed and I am using the latest diag program from WD.

 

On the 500GB drive I let it run for 24 hours just idle to get it good and warm and then I ran Wd's Quick test 5-7 Minutes, then an extended test 2Hrs then I wrote zero's to the drive and finally another extended test and it passed every test with a status code of 0000 - which is no problems found.  I Just finished the 400Gb drive with the exact same results.

 

It was a lot of work to replace one of these drives in the server almost two days as I lost all the data on the 400 and had to rebuild the new drive manually. In all fairness, I was not the most competent at using reiserfsck then, so it could have been some I did in or out of sequence that caused the data loss, but I am wondering now if these drives after 2.5 days of testing a temp have no issues why they were failed in the server.

 

 

Regards,

 

Dave

Sorry you had these problems.

 

I have not (personally) had great luck with WD drives.

 

Losing 2 at once, even with relatively pessimistic reliability estimates, is statistically extremely unlikely.  But we all know that Murphy is alive and well and that no matter what statisitcs say, sh*t happens.

 

The first comment I woud make is that just because unRAID marks a disk as bad, does not mean that thie disk has actually failed.  unRAID will mark a disk bad on a write error.  Disks SHOULD never fail on a write.  It isn't as though the drive rereads what it wrote to make sure it did it correctly.  So a write failure can be quite serious.  But a write failure does not mean the drive is dead, or that you would be unable to retrieve data from it.  I have seen some recently smartctl logs with some read errors that seem to have nothing to do with bad media (no sectors are being marked for remapping even though the read failed).  Similar things could happen on a write.  It doesn't mean the drive is 100% healthy, but likely good enough to get your data off.

 

I have a couple of spare motherboards in my basement and have set up a test unRAID config.  If you set up one (can use free unRAID basic), you could put your "bad" drives in there and try to read from them as disk shares.  If they work, you can (at least) copy the data off. 

 

I had an experience when a neighbor's hard disk failed.  I tried everything to get it working again including putting in the freezer.  He took it to a local repair shop and they were able to get all the data off of it.  No idea what they did.  So if you have important data on a drive, don't give up easily!

 

I posted recently a thread where a suggested burnin method (for two drives) was explained.  It might help burn in your drives in a real-world test using unRAID.  Not sure doing this test would have helped in your situation or not, but couldn't hurt.  http://lime-technology.com/forum/index.php?topic=1887.msg13312#msg13312

 

I recommend that people do parity checks once a month.  If exercises the drive and has benefit far beyond just checking to see if you parity is correct.

 

For similar reasons, I recommend that people do parity checks before opening their unRAID server for any reason, especially if drives are going to be moved around or new ones add.

 

Hope this helps a little.