Quoting Mark Copper <mcopper@titaninterface.com>:

> On Mon, Sep 10, 2007 at 09:46:47AM -0700, tabris wrote:
>> Mark Copper wrote:
>> > Dear Users,
>> >
>> > I have an Intel machine on which I installed software RAID 1 using a
>> > Knoppix trick back in January of last year:
>> >
>> > # uname -a
>> > Linux deneb 2.6.15 #1 SMP PREEMPT Thu Jan 5 18:12:48 EST 2006 i686
>> > GNU/Linux
>> >
>> > The machine suffered occasional kernel panics which, upon removal from
>> > the data center where I colocated it, I have not been able to reproduce..
>> > However, I do notice occasional "hesitations" involving disk writes that
>> > I felt were somehow related to the panics. There was also a post to
>> > kernel.org at the time where in a similar setup kernel panics were
>> > traced to a bad hard disc.
>> >
>> > So, I'm thinking simply to replace both hard drives.
>> >
>> > Is this foolish? Is there a better approach not requiring special
>> > equipment to diagnosing the problem?
>> >
>> > thanks.
>> >
>> > Mark
>> >
>> try smart-tools. a) it can tell the disc to test itself b) it can tell
>> you what the hard-drive thinks about itself (don't pay too much attn to
>> "PASSED" b/c that's just a 24 hour warning)
>>
>> And yes, it does work with SATA drives, it just needs the '-d ata' hint
>
> Thank you for this. My discs get a clean bill of health from SMART.
>
> So I'm left with these hesitations I don't understand. These happen
> with simple bash commands (ls, man, mv) as well as delivery of web
> pages. For instance, I just waited nearly 30 seconds for "man" to
> return, but only when the given command has not been used for a while.
>
> Is there some aspect to disk access that SMART does not test?
>

Have you tried upgrading your kernel to the latest stable release?
2.6.15 is old these days, and if I remember correctly, may have had
some memory bugs.
2.6.18 is Debian's stable release.

Mike