Is it being rebuilt, causing higher IO?:  

http://serendipity.ruwenzori.net/index.php/2007/08/06/mdadm-rebuild20-event-de
tected-on-md-device  

Or could you have replaced a drive at some point and had an alignment
mis-match?  

If they seem "healthy" to mdadm, I would go with the suggestions to check the
smart counters on the drives as others have suggested.  

This document might be a bit old, but might give you some places to look as
well:  

http://tldp.org/HOWTO/Software-RAID-HOWTO-6.html  

Kelly  

   
>  Thu Jul 14 2016 03:37:58 PM CDT from "gregrwm" <tclug1 at whitleymott.net> 
>Subject: Re: [tclug-list] raid corrupted? but no clue from mdstat
>
>    the first report had numbers for all devices, the second and third for
>sda&sdb only, and the 4th and onward are all zeros.  which is probably just
>right for a quiescent box.  but mdadm still has the CPU pinned.
>  
>  
>  On Thu, Jul 14, 2016 at 12:46 PM, kelly <kelly.black at penguinpackets.com>
>wrote:
>  
>>    
>>
>>Does iostat give any clues as to which drive is busy (i.e. more busy than
>>the others as far as % usage?)  
>>
>>iostat -d -x 1  
>>
>>Is the LVM ok otherwise?  
>>
>>Kelly  
>>>  Thu Jul 14 2016 12:32:24 PM CDT from "gregrwm" <tclug1 at whitleymott.net> 
>>>Subject: [tclug-list] raid corrupted? but no clue from mdstat
>>>
>>>        what's up with my mdadm?  note how it's taking 97% of CPU:
>>>....
>>>
>>>  
>>>
>>>  
>>>
>>>  
>>>
>>>  

>>  
>>
>>  

>  
>
>  
>
>  
>
>  
>
>  
>
>   (, 0 bytes) [View| Download]
>   
>
>  
>
>  
>
>  

  

 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mn-linux.org/pipermail/tclug-list/attachments/20160714/106a26c8/attachment.html>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 
URL: <http://mailman.mn-linux.org/pipermail/tclug-list/attachments/20160714/106a26c8/attachment.ksh>