scsi-stop and cron-job?

[Mai Kee Reis]

Can anybody please give me a hint how to spin down my HDD if not used a while?
My system is a NSLU2 with SlugOS 5.3 LE.

The following does work fine, if invoked manually:
/opt/sbin/scsi-stop /dev/sdb

So I think about using cron - good idea?

Or will this cause trouble to attached processes? (Samba,...)
I'm not shure, but I think I saw a solution with crontab + some
kind of disk statistics, but can't google or remember anymore

(invoking '/opt/sbin/scsi-idle /dev/sdb 180' does not do the job
as it shall by documentation)

Greetings,
Mai Kee





------------------------------------

Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/nslu2-linux/

<*> Your email settings:
    Individual Email | Traditional

<*> To change settings online go to:
    http://groups.yahoo.com/group/nslu2-linux/join
    (Yahoo! ID required)

<*> To change settings via email:
    mailto:nslu2-linux-digest@...
    mailto:nslu2-linux-fullfeatured@...

<*> To unsubscribe from this group, send an email to:
    nslu2-linux-unsubscribe@...

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/

[Thomas Reitmayr-2]

Hi Mai Kee,
you can read /sys/block/<dev>/stat (with <dev> corresponding to your
HDD, eg. sdb) and look for changes. After a certain timeout you could
issue your scsi-stop command.
Altough my disk on a TS-109 could spin down automatically, I use such a
script to be able to take other aspects into account - the disk will
never spin down as long as I have one of its NFS shares mounted from my
PC.

That's my script:

=========================================
#!/bin/sh

#DEV=sdb
DEV=`blkid -t LABEL=MEDIA | sed 's/[0-9].*//'`
DEV=${DEV#/dev/}
STATFILE=/sys/block/$DEV/stat
RMTAB="/var/lib/nfs/rmtab"
let 'CYCLES=10*60/5'
LOGFILE=/var/log/spindown.log
state=active

rw=0
count=0
echo "" >> $LOGFILE
date "+%D %T - start" >> $LOGFILE
while true ; do
  rw_old=$rw
  read reads a2 a3 a4 writes rest < $STATFILE
  let "rw=$reads+$writes"
  if [ "$rw" != "$rw_old" -o -s "$RMTAB" ] ; then
    count=0
    state=active
    lastacc="`date '+%D %T - idle'`"
  else
    if [ "$state" = "active" ] ; then
      let "count=$count+1"
    fi
  fi
  if [ "$state" = "active" -a $count -gt $CYCLES ] ; then
    echo "$lastacc" >> $LOGFILE
    date "+%D %T - spindown" >> $LOGFILE
    hdparm -Y /dev/$DEV
    state=standby
  fi
  sleep 5
done
=========================================

You will probably need to tweak it a bit, eg. set DEV=sdb or change the
disk LABEL, remove the code watching $RMTAB, or maybe do not write a log
file (especially if your /var/log directory is not located on a ram
disk). And of course "hdparm -Y" should be your scsi-stop command.
The script will wait for 10 minutes of inactivity, set via CYCLES.

Hope that helps.
-Thomas


Am Dienstag, den 22.09.2009, 19:02 +0200 schrieb Mai Kee Reis:

> Can anybody please give me a hint how to spin down my HDD if not used a while?
> My system is a NSLU2 with SlugOS 5.3 LE.
>
> The following does work fine, if invoked manually:
> /opt/sbin/scsi-stop /dev/sdb
>
> So I think about using cron - good idea?
>
> Or will this cause trouble to attached processes? (Samba,...)
> I'm not shure, but I think I saw a solution with crontab + some
> kind of disk statistics, but can't google or remember anymore
>
> (invoking '/opt/sbin/scsi-idle /dev/sdb 180' does not do the job
> as it shall by documentation)
>
> Greetings,
> Mai Kee
>
>
>
>
>
> ------------------------------------
>
> Yahoo! Groups Links
>
>
>
>

[Mai Kee Reis]

Hi Thomas,

your script helps a lot, I think!
Need to learn sh syntax again ;-) but i like it.
Some small questions:
Do you know what rmtab counterpart for SAMBA is?
Did I understand right, you don't evaluate the in-que-read/write operations?
why can you use hdparm? AFAIK, the slug has no IDE port, and hdparm does
not work via USB, as I read here:
<http://www.nslu2-linux.org/wiki/FAQ/SpinDownUSBHarddisks#method1>

with regards & respect,
Mai Kee

Thomas Reitmayr wrote:

> Hi Mai Kee,
> you can read /sys/block/<dev>/stat (with <dev> corresponding to your
> HDD, eg. sdb) and look for changes. After a certain timeout you could
> issue your scsi-stop command.
> Altough my disk on a TS-109 could spin down automatically, I use such a
> script to be able to take other aspects into account - the disk will
> never spin down as long as I have one of its NFS shares mounted from my
> PC.
>
> That's my script:
>
> =========================================
> #!/bin/sh
>
> #DEV=sdb
> DEV=`blkid -t LABEL=MEDIA | sed 's/[0-9].*//'`
> DEV=${DEV#/dev/}
> STATFILE=/sys/block/$DEV/stat
> RMTAB="/var/lib/nfs/rmtab"
> let 'CYCLES=10*60/5'
> LOGFILE=/var/log/spindown.log
> state=active
>
> rw=0
> count=0
> echo "" >> $LOGFILE
> date "+%D %T - start" >> $LOGFILE
> while true ; do
>   rw_old=$rw
>   read reads a2 a3 a4 writes rest < $STATFILE
>   let "rw=$reads+$writes"
>   if [ "$rw" != "$rw_old" -o -s "$RMTAB" ] ; then
>     count=0
>     state=active
>     lastacc="`date '+%D %T - idle'`"
>   else
>     if [ "$state" = "active" ] ; then
>       let "count=$count+1"
>     fi
>   fi
>   if [ "$state" = "active" -a $count -gt $CYCLES ] ; then
>     echo "$lastacc" >> $LOGFILE
>     date "+%D %T - spindown" >> $LOGFILE
>     hdparm -Y /dev/$DEV
>     state=standby
>   fi
>   sleep 5
> done
> =========================================
>
> You will probably need to tweak it a bit, eg. set DEV=sdb or change the
> disk LABEL, remove the code watching $RMTAB, or maybe do not write a log
> file (especially if your /var/log directory is not located on a ram
> disk). And of course "hdparm -Y" should be your scsi-stop command.
> The script will wait for 10 minutes of inactivity, set via CYCLES.
>
> Hope that helps.
> -Thomas
>
>
> Am Dienstag, den 22.09.2009, 19:02 +0200 schrieb Mai Kee Reis:
>  
>> Can anybody please give me a hint how to spin down my HDD if not used a while?
>> My system is a NSLU2 with SlugOS 5.3 LE.
>>
>> The following does work fine, if invoked manually:
>> /opt/sbin/scsi-stop /dev/sdb
>>
>> So I think about using cron - good idea?
>>
>> Or will this cause trouble to attached processes? (Samba,...)
>> I'm not shure, but I think I saw a solution with crontab + some
>> kind of disk statistics, but can't google or remember anymore
>>
>> (invoking '/opt/sbin/scsi-idle /dev/sdb 180' does not do the job
>> as it shall by documentation)
>>
>> Greetings,
>> Mai Kee
>>
>>
>>
>>
>>
>> ------------------------------------
>>
>> Yahoo! Groups Links
>>
>>
>>
>>
>>    
>
>
>  

------------------------------------

Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/nslu2-linux/

<*> Your email settings:
    Individual Email | Traditional

<*> To change settings online go to:
    http://groups.yahoo.com/group/nslu2-linux/join
    (Yahoo! ID required)

<*> To change settings via email:
    mailto:nslu2-linux-digest@...
    mailto:nslu2-linux-fullfeatured@...

<*> To unsubscribe from this group, send an email to:
    nslu2-linux-unsubscribe@...

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/

[Thomas Reitmayr-2]

Hi Mai Kee,
it is the script I use on my QNAP TS-109, which has its disk(s) attached
via SATA, so for the NSLU2 you would just replace hdparm with scsi-stop.

I am not 100% sure what you mean with "in-que-read/write operations".
The parameters from /sys/block/$DEV/stat used in the script are the
number of completed reads and writes
(see /usr/src/linux/Documentation/iostats.txt). However you could
actually simplify this by just watching for _any_ change
in /sys/block/$DEV/stat, as shown in the following stripped down version
of the script (untested!):

#!/bin/sh

DEV=sdb
TIMEOUTMIN=10
TICKSEC=10
let "CYCLES=$TIMEOUTMIN*60/$TICKSEC"

state=active
stat=""
count=0
while true ; do
  stat_old=$stat
  stat=`cat /sys/block/$DEV/stat`
  if [ "$stat" != "$stat_old" ] ; then
    # disk activity detected
    count=0
    state=active
  else
    [ "$state" = "active" ] && let count++
  fi
  if [ "$state" = "active" -a $count -gt $CYCLES ] ; then
    # spin down
    /opt/sbin/scsi-stop /dev/$DEV
    state=standby
  fi
  sleep $TICKSEC
done

Regarding samba, you could watch the client and/or "Locked files"
entries returned by the tool 'smbstatus' and only spin down the disk if
there are no such entries (if that's the behavior you desire). However I
am not that familiar with samba, so this might not cover all your use
cases...

Best regards,
-Thomas


Am Mittwoch, den 23.09.2009, 23:17 +0200 schrieb Mai Kee Reis:

> Hi Thomas,
>
> your script helps a lot, I think!
> Need to learn sh syntax again ;-) but i like it.
> Some small questions:
> Do you know what rmtab counterpart for SAMBA is?
> Did I understand right, you don't evaluate the in-que-read/write operations?
> why can you use hdparm? AFAIK, the slug has no IDE port, and hdparm does
> not work via USB, as I read here:
> <http://www.nslu2-linux.org/wiki/FAQ/SpinDownUSBHarddisks#method1>
>
> with regards & respect,
> Mai Kee