disk spindown

[Wolfgang S. Rupprecht]

Has anyone succeeded in spinning down disks under Fedora?

I've done the following and the "hdparm -C" does show the disk spun
down, but the next time I look it is spun up, and it stays up.  The disk
is unmounted and I've stopped and disabled smartd, so it isn't accessing
the disks.  I don't think anything else is either.  I'm probably missing
a trick somewhere, but what???

disk=/dev/sdc

hdparm -S 120 $disk
    hdparm -y $disk
    hdparm -C $disk

-wolfgang

[Bill Davidsen]

Wolfgang S. Rupprecht wrote:
> Has anyone succeeded in spinning down disks under Fedora?
>
> I've done the following and the "hdparm -C" does show the disk spun
> down, but the next time I look it is spun up, and it stays up.  The disk
> is unmounted and I've stopped and disabled smartd, so it isn't accessing
> the disks.  I don't think anything else is either.  I'm probably missing
> a trick somewhere, but what???
>  disk=/dev/sdc
>
>  hdparm -S 120 $disk
>  hdparm -y $disk
>  hdparm -C $disk
>
> -wolfgang

The value after -S is not seconds, for people who don't use that option.
The drive will stay up a long time when accessed. As the man page says "The 
encoding of the timeout value is somewhat peculiar." If it doesn't spin down 
after an hour something is wrong.

One thing wrong might be jumpers, Other than that, I suspect something is still 
accessing it. udev?

[Wolfgang S. Rupprecht]

Bill Davidsen  writes:
> Wolfgang S. Rupprecht wrote:
> > I've done the following and the "hdparm -C" does show the disk spun
> > down, but the next time I look it is spun up, and it stays up.  The disk
> One thing wrong might be jumpers, Other than that, I suspect something
> is still accessing it. udev?

Thanks for the good ideas.

This is a WD Caviar Green and the jumpers are only documented to slow
down the SATA by one notch and add spread spectrum clocking for rf noise
reduction.

I'll have to try to check for udev.  Maybe running lsof in a loop will
catch it.

-wolfgang

[Bill Davidsen]

Wolfgang S. Rupprecht wrote:
> Bill Davidsen  writes:
> > Wolfgang S. Rupprecht wrote:
> > > I've done the following and the "hdparm -C" does show the disk spun
> > > down, but the next time I look it is spun up, and it stays up.  The disk
> > One thing wrong might be jumpers, Other than that, I suspect something
> > is still accessing it. udev?
>
> Thanks for the good ideas.
>
> This is a WD Caviar Green and the jumpers are only documented to slow
> down the SATA by one notch and add spread spectrum clocking for rf noise
> reduction.

Wanted to be sure that people knew "-S 120" doesn't mean "spin down after two 
minutes."
> I'll have to try to check for udev.  Maybe running lsof in a loop will
> catch it.

Did you take out the rule for that? I'm not a guru, can't tell you which rule 
without looking it up, but I have found it, since my hot backup drive on a USB 
dongle goes down and stays that way. I bet someone will remind us which rule 
checks that before you can look. ;-)

[poma]

On 20.04.2013 03:34, Wolfgang S. Rupprecht wrote:
> Has anyone succeeded in spinning down disks under Fedora?
>
> I've done the following and the "hdparm -C" does show the disk spun
> down, but the next time I look it is spun up, and it stays up.  The disk
> is unmounted and I've stopped and disabled smartd, so it isn't accessing
> the disks.  I don't think anything else is either.  I'm probably missing
> a trick somewhere, but what???
> disk=/dev/sdc
>
> hdparm -S 120 $disk
> hdparm -y $disk
> hdparm -C $disk

yum info sdparm
…
: Warning: It is possible (but unlikely) to change SCSI disk settings
: such that the disk stops operating or is slowed down. Use with care.

i.e.
/usr/bin/diskungfu:
#!/bin/sh
# Disk stop - spin down

grep -w sdc /proc/diskstats >dstat.1st
sleep 60
grep -w sdc /proc/diskstats >dstat.2nd
if cmp dstat.1st dstat.2nd >/dev/null 2>&1
then
    echo Stopping disk, spinning down…
    sdparm -f -r -q -v -C stop /dev/sdc
    exit 0
else
    echo Disk busy.
    exit 1
fi
EOF

Change the parameters as needed,
set-up a cron job,
and there you go. ;)

poma

[Wolfgang S. Rupprecht]

poma  writes:
> grep -w sdc /proc/diskstats >dstat.1st
> sleep 60
> grep -w sdc /proc/diskstats >dstat.2nd
> if cmp dstat.1st dstat.2nd >/dev/null 2>&1
> then
>     echo Stopping disk, spinning down…
>     sdparm -f -r -q -v -C stop /dev/sdc
>     exit 0
> else
>     echo Disk busy.
>     exit 1
> fi
> EOF

Thanks!  I had forgotten about sdparm. (Boy do I miss the days when I
just installed everything and then could do something like "man -k disk"
to find likely candidates.

Monitoring /proc/diskstats might also give me a clue as to what is
touching the disks.

-wolfgang