Fabien Wernli wrote:

> In my experience, smart isn't quite reliable as it should be:
> I have so many examples where smart says it's ok, but it isn't, or where
> smart says it's dying but the disk works perfectly.
> 
> Thus the only safe way to work it out is:
> 
> 1) if smart says the disk is unhealthy, change it
> 2) if you get I/O errors in syslog, change it
>    even if smart says it's all ok

Interestingly, I think that you are right...that SMART isn't 
perfect...and that
you never really know...I have also had disks that reported errors and keep
running without problems.  Good example is a bad sector, like I have on my
other machine....well..it gets mapped and it won't use that sector 
anymore...

The "long" test is a pretty good one though...and if there are 
problems...I would
say that they would "usually" show up there...the "short" test maybe not...

But....the *only* safe way is to change the disk if there is any 
suspicion...I agree.
Fabien, are there any other things that that can be done to check the 
health of
that disk?  All of the parameters look really good.....


-- 

Those who do not understand Unix are condemned to reinvent it, poorly.
		-- Henry Spencer