14

I use Ubuntu 12.04 on a home server. There are 5 hard drives: one SSD with the OS and 4 HDDs (Samsung HD204UI) configured as a RAID5 using mdadm.

I don't need the RAID very often (say, about once per day in average); it contains multimedia data and some backups, so I decided to put the drives to standby most of the time.

I configured the drives to automatically spin-down after 20 minutes using hdparm, here /dev/sda as an example:

hdparm -S 240 /dev/sda

The output of the command confirmed this:

/dev/sdb:
 setting standby to 240 (20 minutes)

But the problem is that the drive spins down after about 5 seconds (instead 20 minutes) after the last access, and

hdparm -C /dev/sdb

displays:

/dev/sdb:
 drive state is:  standby

Why doesn't the drive wait the 20 minutes as configured? Does it have something to do with the RAID setup? (I don't think so...)

This very short timeout is problematic since it is a RAID5 and when I access a file, it is spread block-wise around the four drives. Since drive 1 isn't needed when drive 2 is read, it went asleep when it will be used the next time (after drive 3 has been read from). So the drives keep spinning up and down the whole time resulting in a throughput near zero.

For the people being concerned about the life of my drives: It will be no problem that the drives are spinned up and down once a day in average since I prefer saving energy and heat over health of the drives (I use RAID5 and the content isn't very important).

leemes
  • 951

3 Answers3

12

Is AHCI enabled in your BIOS? If not try turning it on and re-trying.

If that still doesn't work, then you may have to try something a bit more complicated, as hdparm doesn't always play nicely with SATA drives, which is what I presume you have.

Check out this page for the following relatively simple solution that uses crontab:

sudo apt-get install sg3-utils
crontab -e

Now, add this to the bottom of your crontab (replace sdx with the drive you wish to put in standby):

0-59/15 * * * * ( if [ ! -f /dev/shm/1 ] ; then touch /dev/shm/1 /dev/shm/2; fi ; mv /dev/shm/1 /dev/shm/2; cat /proc/diskstats > /dev/shm/1 ) >/dev/null 2>&1
0-59/15 * * * * ( export HD="sdx "; if [ "$(diff /dev/shm/1 /dev/shm/2 | grep $HD )" =  "" ] ; then /usr/bin/sg_start --stop /dev/$HD; fi ) >/dev/null 2>&1

Of course, before attempting any of this, it's highly advisable to perform a full backup of the contents of your hard drives.

SirCharlo
  • 40,096
1

Maybe a better way could be:

hdparm -B 1 -Y /dev/device
hdparm -B 255 /dev/device

The first hdparm put the device standby immediately. The second one turn off APM, so the disk stay in standby mode till the next disk access.

mat
  • 11
0

I used a systemd unit file to set the spindown timeout to the lowest one 21m15s right after the boot. It's running only once at each boot, because it's just a temporary setting and the drive forgets it. Apart from that my RAID1 is configured as BTRFS filesystem devices and contains a 2TB Samsung 870EVO SSD and a Samsung HD204UI.

/etc/systemd/system/hdparm.service

[Unit]
Description=hdparm sleep


[Service]
Type=oneshot
ExecStart=/usr/bin/hdparm -S 255 /dev/sdb


[Install]
WantedBy=multi-user.target

https://wiki.archlinux.org/title/hdparm#Putting_a_drive_to_sleep_directly_after_boot

vinibali
  • 1