(linux_command_line)-> ide drive power modes (hdparm)
submited by Russell Thu 23 Jun 05|
Edited Wed 15 Feb 06
to check the power status of a drive
hdparm -C /dev/hdd
to tell a drive to go to standby (power saving mode)
drive state is: active/idle
hdparm -y /dev/hdd
hdparm also has a -Y command (sleep) but this seems not as safe and aparently does not save much power. In one example:
issuing standby command
a drive rated at 10.0 watts idle , 1.8 watts standby, 1.3 watts sleep
so the savings for sleep over standby is minor. My main concern in using the stanby state is to reduce heat generated by my stack of 8 drives. Powering down unused drives will help the problem.
I don't know if is linux defaults or hardware defaults that don't seem to put the drives on standby, but that's gotta reduce lifespan gerating all that extra heat.
does power save mode still work with a raid array?:
I am using these settings : (tell all drives to sleep after 5 minutes)
hdparm -S 60 /dev/hd?
On a server that has a raid array. I did this after two drives failed at about the same time. I blamed the failure on heat, so I installed more cooling fans. I got some of these fans that screw to the side of the hard disks, I re-posistioned the drives so that none are right next to each other in the case, and I added the hdparm command to my /etc/rc.local file.
The end result is that I have had no failures, and the only annoyance is that if the server has been idle, you notice a 10 to 20 second delay when it "wakes up" when you go to open a file on it or even get a list of files. The delay is annoying, but tolerable, and you only notice it if the office is empty and you havn't been hitting the server in a while. But it hasn't caused any problems. It will spin-down the drives even if you have a file open on the server. (if you haven't done a read or write to it in 5 minutes) And now, when I check the drives in the server, they are always cold to the touch. ( clearly the fans have something to do with this as well) So I think it's been a good thing that I did this.
But my situation may differ from yours. For one thing, the raid array is not the boot drive. So, any housekeeping work the server does in the background, doesn't touch the array. Also this isn't a web server, it is an office file server that mostly only works during the day. If this was a machine that needed to respond faster more of the time, I don't think I would do this. I also worry that the spin-down and spin-up labor might actualy put more wear and tear on the drives. I figure that if one of the drives just out and out fails, the raid array will cover for it. But what if more than one fails at about the same time? ( because they all have about the same duty cycle) ... Well I have had it this way for about 8 months and not had any problems. If that changes, I will make a note here.
other hdparm options, Force eject (Russell)
Add comment or question...: