Disk chatter every 5 seconds

[Noah Dain]

On 12/18/05, Daniel Allen <dada.da at gmail.com> wrote:
> Thanks. I've checked /var/log/* and nothing's updating at that
> frequency.  I also neglected to mention that I unset APM (hdparm -B
> 255) on the one drive that supported it.  I'm not sure if there are
> other hdparm settings that might be helpful here; googling didn't seem
> to find anything.
>
> As you suggested, I updated /etc/fstab; I ran mount -o remount noatime
> , but no change...
>
> -Daniel
>
> On 12/18/05, Scott J. Henson <scotth at csee.wvu.edu> wrote:
> > Daniel Allen wrote:
> >
> > >Ubuntu (5.04 and 5.10) seem to run a device-related daemon (HALd?
> > >udev?) which makes the disk activate in a short burst exactly every 5
> > >seconds.  Since the disk is the loudest part of my home system, the
> > >pattern is sometimes a bit distracting.
> > >
> > >Anybody know what's the cause, and whether it's configurable behavior?
> > > I have done a bit of poking, but I haven't been able to isolate it,
> > >beyond that it doesn't happen in runlevel 2, and it doesn't stay
> > >active long enough to show up in 'top'.
> > >
> > >
> > That 5 seconds sounds an aweful lot like the 5 second commit interval
> > that ext3fs has.  So its not hald or udev directly.  Its probably
> > something doing some logging or something else.  You may also think your
> > not writting anything to disk, but every you access a file the atime
> > must be updated, so that needs to be written to disk.  On my laptop I
> > set the noatime option on all my partitions.  This will keep it from
> > updating that.  Also you should check in /var/log to see if anything is
> > writting too much to disk and see if you can scilence it.  Generally
> > setting the noatime mount option fixes it.  You can set this in
> > /etc/fstab.  Man mount for a full description of this option.
> >
> > --

try using the commit= mount option to change the write commit times.

also, to see what is accessing the disk:  echo 1 > /proc/sys/vm/block_dump;
and then tail dmesg.  Just don't forget to turn off block_dump when
you are done.

--
Noah Dain
"Single failures can occur for a variety of reasons that have nothing
to do with a hardware defect, such as cosmic radiation ..." - IBM
Thinkpad R40 maintenance manual, page 25