Hitachi HDS7220 Raw Read Error Rate

I have 3 Hitachi HDS7220 2 TB 7200 RPM drives in my array.  One of them is the parity drive.  I was copying over some movies and noticed the drive was making a lot more noise than usual and the write speed was slow.   

I stopped the copy, and ran a smart test on the hitachi data drives. The raw_read_error_rate is climbing significantly each time I run the utility.  The value has dropped from 100 to 95 is 3 or 4 tests and the raw value gone from 0 to 350,000 for both drives.  Both drives still say they passed the test.  The threshold for this value is 16.

The parity drive is just as noisy during the smart test but the value is still 100, and raw value 0.  

So do I instantaneously have two drives going bad?  Anyone else have these drives?

I have 3 Hitachi HDS7220 2 TB 7200 RPM drives in my array.  One of them is the parity drive.  I was copying over some movies and noticed the drive was making a lot more noise than usual and the write speed was slow.  

 

I stopped the copy, and ran a smart test on the hitachi data drives. The raw_read_error_rate is climbing significantly each time I run the utility.  The value has dropped from 100 to 95 is 3 or 4 tests and the raw value gone from 0 to 350,000 for both drives.  Both drives still say they passed the test.  The threshold for this value is 16.

 

The parity drive is just as noisy during the smart test but the value is still 100, and raw value 0.  

 

So do I instantaneously have two drives going bad?  Anyone else have these drives?

Unlikely.  The raw_read_error_rate is not something we have seen correlation with failing drives.  (Primarily you need to keep an eye open for the reallocated and pending sectors.)  There is some evidence that this (and other) fields are vendor specific, and each bit, or groups of bits, have different meanings.  Trying to infer something from the decimal equivalent of a bit mask field is not possible without documentation.  The fact that your value went from 100 to 95 is a bit more evidence of a problem, but with a threshold of 16, it would have to be dropping several times before it starts to look like a trend.  Bottom line, I'd monitor it but wouldn't worry.  There was a video someone posted a while back where a guy screamed into his rack of hard disks and was able to see read errors caused by the vibration of his voice.  If your drives are not level or vibrating excessively, it could cause this type of problem.

I'd be a bit more concerned about the noise.  This is a mechanical issue and may never show up in the smart statistics.  I'd ask some of the same questions about whether the disks are mounted well or vibrating.  Make sure they are level (either vertical or horizontal) and not skewed a little.  Also check the temperature.  If they are running hot I'd be worried about that.

I remember a long while back someone posted a link to audio recordings of common drive sounds indicative of different specific problems.  Maybe you can find that site and diagnose your sound!

Otherwise, I would keep an eye on this drive.  Run some speed benchmarks, parity checks, long drive tests.  If it gets worse or there are signs of performance or smart issues, RMA it.  

Unlikely.  The raw_read_error_rate is not something we have seen correlation with failing drives.  (Primarily you need to keep an eye open for the reallocated and pending sectors.)  There is some evidence that this (and other) fields are vendor specific, and each bit, or groups of bits, have different meanings.  Trying to infer something from the decimal equivalent of a bit mask field is not possible without documentation.  The fact that your value went from 100 to 95 is a bit more evidence of a problem, but with a threshold of 16, it would have to be dropping several times before it starts to look like a trend.  Bottom line, I'd monitor it but wouldn't worry.  There was a video someone posted a while back where a guy screamed into his rack of hard disks and was able to see read errors caused by the vibration of his voice.  If your drives are not level or vibrating excessively, it could cause this type of problem.

 

I'd be a bit more concerned about the noise.  This is a mechanical issue and may never show up in the smart statistics.  I'd ask some of the same questions about whether the disks are mounted well or vibrating.  Make sure they are level (either vertical or horizontal) and not skewed a little.  Also check the temperature.  If they are running hot I'd be worried about that.

 

I remember a long while back someone posted a link to audio recordings of common drive sounds indicative of different specific problems.  Maybe you can find that site and diagnose your sound!

 

Otherwise, I would keep an eye on this drive.  Run some speed benchmarks, parity checks, long drive tests.  If it gets worse or there are signs of performance or smart issues, RMA it.  

Thanks for the reply.

The drives are mounted in my server off the floor, and are horizontal.  They are stable, mounted with all screws and kept continuously cool <35C when in use.  There have not been any vibrations at all, they aren't next to any speakers or home stereo equipment.  I honestly can't imagine anything in this environment that could be that type of issue.

What worries me is two of these drives doing this simultaneously and one of them not doing it at all (though still being a little noisier than I recall).  The two drives that are showing this, were bought at the same time and have ~10-11,000 power on hours.  They are about 18 months old.  The parity drive is a little newer.

I re-ran the tests, and they are slower getting better now.  values coming up and raw values going down.  Still making same noise.

not sure what to make of it, values are up and down in that same range.  copy speeds are more normal on a few runs. 

I think i'll just watch it, may call hitachi and see what they think on Monday.