Problem with Indigo 2

John Andrew Shepard schrieb:

You may suffer from disk stiction.

Here are some previous postings on this subject:

- Disk Stiction is the head sticking to a disk platter. IMHO this
is a problem on all IBM mfg. hard drives from SGI. I have been told
it is only a problem on newer drives, not so. We have some IBM
drive arrays on one of the first rack mount onyx's made and it has
such bad stiction problems that I really try to never shut down the
arrays...Never almost. In my experence patch 466 is a slight help
but nothing to write home about. If you shut down a drive (IBM) and
only get funny noises when you reboot you might have stiction.
Don't laugh... Take the drive out and pretend you are throwing a
frisbee (only don't let go). Do this a few times intersperced with
hitting the top of the drive (and random cursing) reinstall and hope
for the best.

- We've told all sgi support folks worldwide how to do deal with this,
through several different channels, and a "shock" to the drive is
expressly strongly advised against, because it can permanently trash
the drive.

- Personally, I've been told two different "blessed fixes" for the
drive stiction problem -- 1) "hold drive like frisbee, throw like
frisbee, only don't let go"; and 2) "hold drive on edges like pack
of cards flat against chest, push arm outwards and back very
sharply".

Only 1 is blessed by IBM's disk folks, and by SGI. 2 is better than
dropping the drive, but can still trash the hda. As I said in an
earlier reply tonight, when the drive is stuck, the data can still
be recovered, almost all the time. But not after the HDA is
damaged.

- The only SGI/IBM approved way is `throw it like a frisbee, but don't
let go' or anything that causes the same movement. The idea is to
have the disk platters move in their normal direction relative to
the heads.

DON'T hit it, drop it, or otherwise do anything that forces the
heads to bounce off the surface. This might make it work, but you'll
have a *much* greater risk of breaking things.

After the disk is unstuck, it's wise to check for the presence on
the system and in the crontab of patch466, aka disk stiction patch.

Alternative: call an SGI SSE. He should be experienced with this:
we've spent some time with the UltraStars before patch466 went out.

- If it is a IBM DFHSS disk and if it is stuck you will hear a beeping
sound at power-up.

In this case the patch won't help. Twisting the disk unit abruptly
around the axis of the drive helps.

I have applied this technique successfully more than a dozen of
times but after some time I replaced these unacceptable disks by
another model.

- Having heard all this discussion about stiction and disk drives,
how does it manifest itself? My three SGI macines stay on, and
so power cylces on the drives are not a ususla occurance. Once
again, how do I know which drives ar vulnerable, and what are my
options?

If they are IBM Ultratar drives you might be at risk.

scsicontrol (from my web page) will show an inquiry string
similar to this:

canto 1# scsicontrol -i sc1d1l0
sc1d1l0: Disk SGI IBM DFHSS2E 4C43

Or if you have patch 466 installed:

canto 2# diskpatch -v
sc1d1l0: Disk SGI IBM DFHSS2E 4C43
Date: 95/023 Serial: 00082569

(The date code is only shown in the 6.2 version of diskpatch; this
happens to be one of the drives in the affected datecode range.)

- Having heard all this discussion about stiction and disk drives,
how does it manifest itself? My three SGI macines stay on, and so
power cylces on the drives are not a ususla occurance. Once again,
how do I know which drives ar vulnerable, and what are my options?

There's a script available that will check for you at the Silicon
Surf web site #

- Earlier this year SGI replaced all the 2 gb IBM drives in our
Indigo2s because of stiction. This was done under a special recall
program.

Last week we had to shut off 20 of our systems so PG&E could do some
work. When we turned them back on, 6 of them failed to come up
because of stiction. Fortunately gently giggling the drives unstuck
them and I immediately reinstalled patch 466 on the machines that
were still running 5.3.

Has anyone else had any problems with the replacement IBM drives? I
did ask for Quantum drives when they were first replaced. Our F.E.
assured us though that IBM had fixed the problem and the new drives
wouldn't stick.

- That's *all batches* of the ultrastar. The risk is greater (much
greater) on some batches, but all of the ones produced to date have
measurable risk of stiction. The patch drops that down into the
background for most of those drives, if actually used.

- They may not be dead! Search the net for the word "stiction".
It's a common problem in older drives. You *might* be able to
get them to spin up by taking them out of the enclosure, holding
them in your hands and quickly "jerking" them about the axis
of the platter spindles. Think of it as trying to move the
platters inside the drive. Be careful to hold the drive
securely while you're doing this so it doesn't fly out of your
hands. Probably wouldn't hurt the drive, but you might kill
your cat or take a large chunk out of your hardwood floor.

It's caused by the bearings getting a buildup of something or
other inside. The spindle motor doesn't have much torque
at startup at all (that's why they take so long to spin up)

Note that a common misconception about stiction is that it's
caused by the drive heads on the platters. That's not the case,
as all drives since the days of the Seagate ST-225 and Miniscribe
3425 have auto parking heads. Uh oh. Am I showing my age here?

I prefered the following method: put the disk on a piece of cloth
on a horizontal flat surface (table). Then turn the disk a few
times rapidly around its vertical axis (i.e. around the spindle).

Regards
Ralf Beyer
--