Date: Tue, 17 Jul 2001 00:37:47 -0400 From: Bob Johnson <bobj@ufl.edu> To: "Gray, David" <David_W_Gray@tvratings.com> Cc: "'FreeBSD mobile list'" <freebsd-mobile@FreeBSD.ORG>, blovett@bsdguru.com Subject: Re: Disk clicking... (Was: Re: Dell Inspiron 8000 and suspend-to-disk) Message-ID: <3B53C11B.1E692E47@ufl.edu> References: <01D4D419B1A4D111A30400805FE65B13070AC38F@nmrusdunsx1.nielsenmedia.com>
next in thread | previous in thread | raw e-mail | index | archive | help
"Gray, David" wrote:
>
> I have an older Toshiba Tecra with a 5G IBM drive that does this. It
> runs *extremely* slowly under load. It appears to be re-calibrating
> (thats what a seek to home position, then back to the target track
> is called) constantly. This is related to the ATA driver. If I use
> the WD driver (deprecated), it works just fine. Or, {shudder}, Win95.
>
Well, you got me curious, so I started paying close attention to
what my Inspiron 7500 is doing. I even booted Windows (the things
I do for FreeBSD...) and used it for a while.
So here's what I think I am hearing:
1) Seeks are a fairly noisy operation on this drive. Activities
that only require track-to-track seeks (or at least short seeks)
are quiet, e.g. deleting /usr/obj or booting FreeBSD in the first
place. Things that move the heads long distances are noisy.
2) The drive parks after 8 seconds (roughly) of inactivity. This
operation is fairly noisy and takes about 1/2 second. It is
distinguished by the fact that the disk activity light does not
come on when you hear the activity. The park operation probably
includes a calibration (that would help explain why it takes so
long).
3) Because the heads are parked, any new activity begins with an
initial seek that covers enough distance to be fairly noisy.
It takes long enough that I suspect it includes a calibration.
4) There is not noticable difference in disk activity or the amount
of disk noise that I hear in Windows and FreeBSD. There is enough
ram in the system that neither needs to be doing swapping. Almost
any activity in Netscape causes a noisy disk operation (this may
be more prominent in FreeBSD, but if so I think it is a function
of the application, not the operating system).
5) The noises are present even in the early stages of a cold boot,
when the BIOS is still controlling the system. My system has a
boot password set, and when it stops to wait for that password,
I hear the drive park, for instance.
6) The sound of the spinning disk seems to be getting louder
with time. I hope that's my imagination, because I consider it
to be far more sinister than noisy seek operations. It is a sign
of failing bearings. Maybe I should see if I can get Dell to
replace it before it fails, instead of waiting for it to do so.
- Bob
> I have a brand new 20G to go in the thing - we'll see if that works
> any better. Otherwise, it looks like 4.2 is the end of the line for
> this fella. (I *did* see that WD was axed completely nowadays, didn't
> I?)
>
> -----Original Message-----
> From: Ben Lovett [mailto:blovett@bsdguru.com]
> Sent: Sunday, July 15, 2001 10:40 AM
> To: Bob Johnson
> Cc: mobile@freebsd.org
> Subject: Re: Disk clicking... (Was: Re: Dell Inspiron 8000 and
> suspend-to-disk)
>
> Bob Johnson (bobj@ufl.edu) wrote:
> > Ben Lovett wrote:
> > >
> > > I believe I saw Greg Lehey (grog@FreeBSD.ORG) write this:
> > > > On Wednesday, 4 July 2001 at 21:05:28 -0700, Ben Lovett wrote:
> > > > > I believe I saw Bob Johnson (bobj@ufl.edu) write this:
> > > > >> Nick Sayer wrote:
> > > > >>> Ben Lovett wrote:
> > > > >>
> > [...]
> > > > > I believe this is the reason behind some strange noises that I hear
> > > > > comming from my disk every once in a while. It sounds like one of
> > > > > the heads is moving rather abruptly, or something like that.
> > > >
> > > > Well, the heads all move together.
> > > ok..
> > > >
> > > > > Has anyone else noticed similar behaviour on Dell i8k's built as of
> > > > > late?
> > > >
> > > > A lot of us noticed "clunk" noises from the disks of 7500s built
> > > > between 12 and 18 months ago. I had mine replaced as a result, and
> > > > the new disk has never made any noise. Mike Smith didn't have his
> > > > replaced, and the disk still goes "clunk" after 18 months, but he
> > > > doesn't have any other problems with it.
> > >
> >
> > My 7500 makes what you might call "clicks" or "clunks" when it does
> > what sound to me like long seeks. I've heard it from so many drives
> > that I consider it normal. My mother-in-law's laptop had a horribly
> > noisy and slow drive that she thought was failing -- it was merely a
> > highly fragmented Windows swap file. Cleaned that up and the system
> > got quiet and fast again (isn't that neat? Windows is actually
> > designed to get slower with use!).
> >
> > When doing most disk-intensive operations on this 7500 (e.g. deleting
> > /usr/obj) it is very quiet, but during normal operation it clicks
> > and clunks once in a while. I've always assumed that was a long seek
> > related to flushing a buffer or something. Long seeks tend to be
> > (in my experience) relatively noisy on many drives.
>
> Yes.. This is what I seem to notice. The funny thing is, I don't
> believe i've heard it when the system is booted into Windows..
> >
> > > Well.. I guess that I am observing this behavior on my replacement disk,
> > > I'll live with it. Unless I hear different.
> > >
> > > On a sort of side note.. I was discussing this issue with my boss
> > > yesterday, and he was explaining to me that when hard drives are
> > > manufactured, because of the nature of the mass-production, the platters
> > > will never turn out without flaws, and thus will have some bad sectors.
> > > So, when the drives go through their low-level formatting at the
> > > factory, those sectors will be marked as "bad" on some part of the
> > > drive. He was telling me that the reason that the drive may be clicking
> > > is that the heads are comming across one of these sectors that is marked
> > > bad, when it thought there was data there (or something to that effect),
> > > and then resyncing itself. Does that make sense? I believe that is
> > > what he was saying.
> >
> > A sector mapped bad shouldn't be accessed in the first place, but
> > it sort of makes sense. My understanding is that when a SCSI drive
> > (and probably recent IDE drives) detects errors in a sector, it maps
> > out that sector and replaces it with one from a spare sector pool at
> > the end of the disk. Thus, when that sector is accessed, there is a
> > seek to the end of the disk which you might hear. I assume that
> > sectors mapped bad at the factory are replaced by the next available
> > sector, so there would be no seek involved. Thus, only newly bad sectors
> > would cause noise.
>
> Yes. That is what I remember now.. He mentioned something about a spare
> sector pool, but I had forgotten about that until now. But, with a
> brand new drive, there should be no newly created bad sectors...
> Atleast not *yet*
>
> -ben
> >
> >
> > >
> > > Thanks again..
> > >
> > > -ben
> > > --
> > > -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
> > > Ben Lovett printf("Hello world!);
> > > blovett@bsdguru.com return 0;
> > > -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
> > > If Bill Gates had a dime for every time a Windows box crashed...
> > > ...Oh, wait a minute, he already does.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-mobile" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3B53C11B.1E692E47>