Re: noflushd: making/having drives spin down.
On Tue, Jan 06, 2004 at 02:44:01PM +1100, Tim Connors wrote:
> On Tue, 6 Jan 2004, Micha Feigin wrote:
> > On Mon, Jan 05, 2004 at 01:36:48PM -0800, Mike Mestnik wrote:
> > > I have a dell8900 and I can't seam to get the drive to spin down on 2.6.0-test9. The drive makes
> > > a noticeable whine that can be heard across the room. I'd like the noise to go away, that is more
> > > important than saving battery power for me.
> > >
> > > I'm using noflushd but it has no effect, I also tried hdtune.
> > >
> > Look at the -M option of hdparm:
> > -M get/set acoustic management (0-254, 128: quiet, 254: fast) (EXPERIMENTAL)
> That'll do no good.
> -M just makes seeks slower. The OP is complaining about the constant whine
> of the HD spinning. I would agree with him - my I4000 is really quite
> disturbingly loud -- in a room full of computers.
It makes seeks slower in order to make the hard disk quieter (I think it
> noflushd is the best solution, as long as you don't use journalling
> filesystems - although there is currently work in that area (seek google
> for "laptop-mode" kernel patches)
I am using laptop mode and its quite good for saving battery power. As
for what would prove better for you is a question of your hard disk
laptop-mode (and probably noflushd) delay writes to disk so that it can
completely spin down, thus if you only occasionally access the disk for
short whiles the -M flag will give you better performance since the
hard disk won't have to spin up again each time you access the disk. If
you do a lot of short reads then laptop-mode and noflushd will do no
good anyway. If you do long reads every once in a while then
laptop-mode and noflushd will be a better solution.
> Having said that, noflushd induces a bug in the kernel that is a killer
> for me - threads become zombied after 43 minutes, eventually filling the
> process table:
> (gee, that mailing list has a lot of spam)
> TimC -- http://astronomy.swin.edu.au/staff/tconnors/
> The Klein-Gordon equation was derived by Schroedinger.
> Hence its name. -- Peter Robinson, Rel. Quant. Mech Lecturer.
> To UNSUBSCRIBE, email to email@example.com
> with a subject of "unsubscribe". Trouble? Contact firstname.lastname@example.org