Re: SIDER'S

[Dan Foster <usenet@evilphb.org>]

In article <20040706021531.04264.00001026@mb-m15.aol.com>, Michael J. Mahon <mjmahon@aol.com> wrote:
> Wayne Stewart wrote:
>
><snip>
>
>>I've wondered whether a drive should be
>>spun up occasionally if stored long term or if it matters at all.
>
> I wonder, too.
>
> If the motor drives the platters using an elastomer belt of
> some type, then spinning it up occasionally so that the
> belt comes to rest in a different position certainly seems
> like a good idea.  (This is certainly the case for floppy
> drives.)
>
> If the motor is direct drive, then perhaps spinning it up
> at least redistributes the lubricant in the bearings, which
> sounds like a good thing.
>
> Then, of course, there are the electrical advantages of
> keeping the electrolytics formed, so they work when you
> need them.

I'd agree with all of the above. I've got a setup that I adminster with
some thousands of hard drives; most used 24x7x365 but a few used less,
and even fewer powered off for long durations.

In my experience, yes, it is indeed imperative to use drives with some
regularity for the various reasons that Michael cited... and at the
other end of the range, it is also good to reboot occasionally because
sometimes you find out a drive has developed an otherwise undetectable
mechanical fault rendering it unusable. (People doing a rare reboot for
Y2K fixes in late 1999 sometimes ran into this, not expecting it!)

Drives in *general* are not very tolerant of just idly sitting on a
shelf somewhere for years on end. Some are more tolerant than others,
but ultimately, physical factors will influence its ability to spin up.

Chemical changes, electrical, other physical changes, dust and
lubrication, moisture, heat, physical jarring motion (e.g. handling),
static electricity, amongst others all influences its ability to
correctly function after an extended period of disuse.

Stiction is the simplest of all disk related issues for anyone without
tools to resolve although it has potential to damage disks. (I have
recovered a few drives this way, done carefully and in the right spots.)
Most other failure modes are not such so simple to diagnose and fix.

It would not be a bad idea to run a drive through a random set of reads
with a reasonable distribution throughout the whole drive (or even a
full-drive block-by-block read) for at least once a week if possible.

(This is how server class disk arrays do it, by running reads across all
disks once every few days to detect any early warnings of potentially
bad trends, for instance, along with monitoring the disk and
controller's reported errors.)

Sure, it's an hassle; how important is preserving the data to you? Most
people don't have the expertise to debug and fix low-level drive
mechanical failures, so it's often prudent to keep it in good shape.

Also, reads and checksumming will more likely tell you as an early
indication when the media surface is starting to fail. They can 'fail'
due to changes in magnetic properties (try to keep them away from large
electromagnets like TVs, large speakers, or monitors... and keep the
drive well shielded) or due to chemical changes such as oxidization.

It's a lot easier to find out something is on its way out and being able
to transfer its data off to something else while it's still accessible
than the utter shock of not seeing it boot one day or be accessible at
all.

-Dan