OT: server maintenance

[Fairlight]

When asked his whereabouts on Fri, Nov 05, 2004 at 07:21:59PM -0500,
Bill Vermillion took the fifth, drank it, and then slurred:
> 
> > Problem is, AMD's don't have step-down.  If they overheat, they keep
> > running at the same rate.
> 
> That was only on the Athlon - not all AMDs.

I thought Duron as well.  No?

> > Your real iNTEL's will scale back, so that P4-3.8GHz might
> > actually be running at 700MHz after a while--but it hasn't
> > fried.
> 
> That's only on the P4s.

I was told P3 and up.  

> > came out that had the huge heat issues (1000 and 1800 Athlons,
> > I remember) were problem children--but then they came out with
> > mainboards with BIOSes that would monitor the chip and shut the
> > system down (or sound an alarm or something) if it overheated.
> 
> They did NOT sound an alarm.  I believe Tom's Hardware tested
> on and the time frame was less than 2 seconds from fan failure
> until the chip melted and charred the motherboard.

Sure--if they were unmonitored.  But that was before they came out with
BIOSes and board that monitored and shut it down.  The Athlons remind me
of Quantum Fireballs--aptly named, apparently.  Someone said that if you
didn't have 'x' amount of 'proper' airflow, they'd last about five minutes
before burning out.  (Why do I -know- you're going to tell the Barraccuda
documented precise orientation/airflow anecdote in response?)  :)

> > (Just what you want--a system shutting down arbitrarily.)
> 
> Better than setting the house on fire.

Well, there -is- that.  :)  

> > This first and only AMD I owned was a 486/120, and I apparently
> > got lucky because I never had problems with mine, but most
> > of that line were supposed to be fickle as you'd ever think
> > something could get. That hasn't been powered up in six years
> > or so though.
> 
> I had no problems with the K3s.  I only undid the system in the end
> as it appears that some regulaors on the motheboard didn't take to
> kindly to a series of power failures, surges, close lighting
> strikes, and things such as CD-ROMS started running very hot.
> That'w why I surmise it was a bad/weakend VR.

You know what, I told a lie--not intentionally though.  I recently was
graciously given a K6-450 specialised linux appliance.  Until you mentioned
the K series, I'd entirely forgotten that I have a K6 in commission.  A
fairly recent addition--and it does what it does very well, I might add.

> They were the best of the Intel work-alikes.  The Cyrix and Blue ??
> from IBM - just weren't as compatible as they should be once you
> got outside of the pure MS world from my observations.

Cyrix had -so- many problems.  When you see things like "cyrix bug" listed
in /proc/cpuinfo checks, you know you just might want to stay away from them.

> > Had one of the longest lived HD's I've had in it though...an
> > IDE Maxtor that ran from 1992 through 1999 and was still
> > showing no signs of failure whatsoever. My Seagate 'cudda's are
> > starting to die...at least one of them...a very slow death.
> 
> Which 'cuda.  SCSI or EIDE.  I've had early 9GB 'cuddas that got a
> bad rep for dying prematurely run for in excess of 6 years in
> 24x7 in a mail server.  The trick to keeping them alive was
> cooling.

2.2GB SCSI.  A twin pair of ST32272N's.  The two that -finally- worked
after going through ***twelve*** bad drives from them, including eight bad
Medalist Pro 2170's and four 'ccuddas of a different model and .1 less
gigabytes of capacity.  That experience is why I've bought anything -but-
Seagate ever since.  My experience with them goes back to someone that had
a Wren VI that lasted for 9 years, through what I went through.  The
general impression I get is that if they make it to you without being DOA
-and- make it through the first 30 days without dying, they'll last 2-3
times their rated MTBF.  But the DOA/30-day-death ratio is incredibly high
from my observations.

> You probably won't hear that in modern drives as they have been
> been improved so much mechanically.

How "modern"?  Seagate Medalist Pro SCSI 2170's do it.  Granted, that's
probably a six or seven year old model, but...  Seriously, -how- modern?  :)

(EVERY Fujitsu I've had went through that death, btw.)

> > That's what usually goes on me is the drive motor bearings.
> > And that can take up to 18+ months once it starts doing it
> > seriously.
> 
> I haven't heard that on any drive since the old 1/2 height [not the
> 1" height] cuddas.

I have, although they were on lesser Seagates.  I'm not sure which of my
drives is starting to do it, either.  I've hacked scsi-idle to work against
the 2.4 kernel, so I can spin them down one at a time, but every time I go
to do it, it stops doing it before I can actually get rebooted to linux and
test to see which drive will die. *grumble* It could be the Conner 219MB
(31200?) I have in there--bought it used when that was adequate swap space,
and it came out of a SparcStation 20.  I don't think it's my IBM UltraStor
9.1GB though.  The Seagates are anyone's guess.  Like I said, it's not
doing it persistantly enough to be tested--yet.  Give it time. :-/

mark->
-- 
Bring the web-enabling power of OneGate to -your- filePro applications today!

Try the live filePro-based, OneGate-enabled demo at the following URL:
               http://www2.onnik.com/~fairlite/flfssindex.html