Hi,

 

I am trying to figure out a way to monitor my hard-drive spin-downs/-ups for a longer period of time, say a week or two. At the moment I have the feeling that my drives are almost always all active causing pretty high power consumption, but it is only a feeling, I would like some real data for this and the log output is kind of hard to follow (lots of spindown messages in there, but I can't really tell for how long and when they woke up without guessing. Is there a way to monitor the spindown state without waking the drives up ? A very simple bash script that monitors the relevant disk state and logs the status every 15 minutes to a file would be perfectly sufficient, I don't know if that is possible though without waking up the drives ?

 

Thanks!

You can monitor last IO time by checking the output of

/root/mdcmd status

 

I use the following to determine the last_io time of my disks.  you can adapt to log every 15 minutes or so if you like.

#!/bin/bash
#Assume inactive if no read/write in 900 seconds if $1 arg is not given.

if [ "$1" = "-q" ]
then
  quiet_mode=yes
  shift
else
  quiet_mode=no
fi

if [ "$1" = "-a" ]
then
   verbose_mode=yes
   shift
else
   verbose_mode=no
fi

no_activity_time=${1-900}

#Get the disk spinup status
disk_status=`/root/mdcmd status | strings | grep rdevLastIO`
# Get the current time.
now=`date +%s`
longest_elapsed_time=0
most_recent_disk_activity=100000000
most_recent_disk=""

function hms() {
    seconds=$1
    hours=$(($seconds / 3600))
    seconds=$(($seconds - ($hours * 3600)))
    minutes=$(($seconds / 60))
    seconds=$(($seconds - ($minutes * 60)))
    [ "$hours" = "0" ] && hours="" || hours="$hours hours, "
    [ "$minutes" = "0" ] && minutes="" || minutes="$minutes minutes, "
    echo $hours$minutes$seconds
}

for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
do
    last_access=`echo "$disk_status" | grep "rdevLastIO.$i=" | cut -d"=" -f2`
    if [ "$last_access" = "" ]
    then
      continue
    fi

    elapsed_time=$(expr $now - $last_access)
    i=`printf "disk%d" $i`

    [ $verbose_mode = yes ] && echo "Most recent disk I/O was on $i `hms $elapsed_time` seconds ago."
    if [ "$last_access" != "0" ]
    then
       if [ ${elapsed_time} -le ${most_recent_disk_activity} ]
       then
          most_recent_disk_activity=${elapsed_time}
          most_recent_disk="$i"
       fi
       if [ ${elapsed_time} -gt ${longest_elapsed_time} ]
       then
          longest_elapsed_time=${elapsed_time}
       fi
    fi
done

if [ $longest_elapsed_time = 0 ]
then
    [ $quiet_mode = no ] && echo "No disks are spinning."
    exit 0
else
    if [ $most_recent_disk_activity -gt $no_activity_time ]
    then
        [ $quiet_mode = no ] && echo -n "All disks inactive for at least `hms $no_activity_time` seconds. "
        [ $quiet_mode = no ] && echo "Most recent disk I/O was on $most_recent_disk `hms $most_recent_disk_activity` seconds ago."
        exit 0
    else
        [ $quiet_mode = no ] && echo "Most recent disk I/O was on $most_recent_disk `hms $most_recent_disk_activity` seconds ago."
        exit 1
    fi
fi

 

invoked as

last_io.sh

it will print the most recent IO time.

invoked as

last_io.sh -a

it will tell you if which, if any disk has been active within the past 900 seconds. (15 minutes)

the exit status will be 0 if all idle within 900 seconds, even if still spinning. 

The exit status will be 1 if any activity within 900 seconds.

 

You can change the time test by adding an argument to the command

last_io.sh 600

will check if there was activity within 600 seconds.

 

You can show all the disks individually with

last_io.sh -a

or

last_io.sh -a 600

 

you can make it completely silent by using the -q option

last_io.sh -q

or this time checking for activity within 1800 seconds.

last_io.sh -q 1800

 

When I experimented with a cache drive I used this to invoke the mover script only when the disks have not been written/read  from the past 15 minutes.

 

In my cron entry for the mover script I used a simplified version named "are_disks_idle.sh"

The cron entry looked like this:

*/5 * * * * /boot/are_disks_idle.sh && /usr/local/sbin/mover 2>&1 | logger

#!/bin/bash
###############################################
# are_disks_idle.sh
# exit status = 0 if all disks idle
# exit status = 1 if any disk spinning
# exit status = 2 if array is not atarted
#
# April 2010 Joe L.

# Check if the array is started
if [ -d /mnt/cache0 ]
then
  # rdevLastIO will be non-zero if a disk is spinning
  last=`/root/mdcmd status | grep -a rdevLastIO | grep -v '=0'`
  if [ "$last}" = "" ]
  then
    # all disks are idle
    exit 0
  else
    # all disks are not idle
    exit 1
  fi
else
  # array is not started
  exit 2
fi

 

 

have fun.  Since unRAID is tracking the last IO to a given disk, you do not need to spin them up to see if they have been active lately.

 

Interesting, thanks alot !

 

This will probably replace the method I found while googling for some time this afternoon, but for the sake of providing an alternative solution I will post what I have found:

 

This comes from this blog post here: http://blog.th-neumeier.de/2011/03/suspend-idle-harddisks-and-monitor-the-suspension/

 

The script on that webpage uses "smartctl -a -n standby /dev/sdX" to query the state of the harddisk without spinning up the drives. It then allows writing the result of that query to a sqlite database and generates a report from that on demand.

 

The cronjob entry looks like this:

*/1 * * * * /boot/extra/disk-suspend-state.sh1 collect 1>/dev/null 2>&1

 

And a

/boot/extra/disk-suspend-state.sh1 generate

 

Gives me a detailes report of what has happened in the last day.

Here is an example output, that also details why I will probably switch to your solution:

 

sdg:

4541 - min Start Stop Count

4542 - max Start Stop Count

1 - difference

States:

97 - active/idle

216 - standby

sdh:

- min Start Stop Count

- max Start Stop Count

- - difference

States:

314 - standby

sdi:

1490 - min Start Stop Count

1490 - max Start Stop Count

0 - difference

States:

60 - active/idle

254 - standby

 

As you can see, three different drives (they are different drive types) report different SMART values, making things a bit hairy.

 

The start-stop-count values seem very unreliable, either they aren't reported at all or they don't change even if there is a change. What worked is the report for the drive states. The drives are monitored every minute, and as you can see that drive /dev/sdi for example was on for around 60 minutes (exactly what I have set for the idle time until drive powerdown) and then in standby for the rest of the time. For /dev/sdg this differed, which is reasonable since there was activity on my server that mgiht have caused a read from the drive.

 

Thank you very much for posting your script though: it works without a sqlite database and it should give me all the info I need, so this will probably replace what I have!