"ad0: TIMEOUT - WRITE_DMA" type errors with 7.0-RC1

[Chuck Swiger]

On Jan 25, 2008, at 1:05 PM, Thomas Hurst wrote:
>> These numbers are quite worrysome-- they should be zero or nearly  
>> so in a
>> healthy drive.
>
> No, these are perfectly reasonable for a Seagate.  I have about 12
> 7200.X's and all show the same sort of behavior.  If they're nearly  
> zero
> it's probably a sign your manufacturer isn't actually counting them
> (marketroids hate accurate SMART readings).
>
> Try graphing them as counters; with an idle disk you'll see periodic
> sawtooth patterns as the heads crawl from one side of the disk to the
> other.

SMART attributes which end with _Ct or _Count are supposed to  
increment with every event; things which end with _Rate (ie,  
Raw_Read_Error_Rate, Seek_Error_Rate) are supposed to indicate the  
frequency of such errors over time.  It would be reasonable for  
Hardware_ECC_Recovered to keep the incremental count, but not the  
other two.

I agree that minor periodic errors happen over time and are not a  
great concern, but a happy drive will show zero reallocated sectors,  
or perhaps a few over the span of a year or two, and will have a ECC  
recovered or UDMA_CRC count which is much smaller than was reported by  
Joe.

YMMV, of course...

-- 
-Chuck