Hardy hdparm does not honour power management options, leading to premature disk failure

[Ashley Hooper]

Binary package hint: hdparm

I have parallel installs of Gutsy & Hardy on my laptop, and some time ago read about the load cycle count problem and how to stop the rapid increase in load cycles by disabling power management via hdparm. I immediately configured hdparm.conf and installed smartmon-tools to watch the count, and it seemed to slow the increase. At that point I was on approximately 130,000 cycles on a 6 month old hard disk.

In recent months I've been using the Hardy Heron alpha, and on installation I configured hdparm and smartmon-tools in the same fashion as with Gutsy. I recently happened to check the count, and noticed to my alarm that my disk is now over 170,000 load cycles (disk failure becomes increasingly likely around 200,000 cycles).

I immediately checked my hdparm.conf etc, noted everything was correct and then ran hdparm from the command line as follows:

 sudo hdparm -B 255 -S 240 /dev/hda

(-B) setting the APM setting to 255, thereby disabling APM on the drive altogether
(-S) setting spindown_time to 240 (i.e. after only spin down after 20 mins of inactivity

I then ran:

 smartctl -a /dev/hda | grep Load_Cycle_Count

to watch the counter. In spite of the hdparm command I'd given, load cycles continued to accumulate at the rate of several per minute (I could hear the drive quietly clunking each time).

I can not continue to use Hardy until this problem is rectified.

[_oOMOo_ (hermann-blaxhall)]

This is probably not your issue, but just for reference Hardy refers to my hard drive as /dev/sda where Gutsy referred to it as /dev/hda - and I had to alter hdparm.conf accordingly for it to have the desired effect.

[Ashley Hooper (ash-hooper)]

Thanks _oOMOo_ - I had noticed that with the various updates to Hardy, the disk device was swapping between sda & hda, and for that reason added stanzas for both disk in /etc/hdparm.conf.

Strangely, I'm unable to reproduce this fault right now but will update the ticket if/when I see the problem again.

[ethanay (ethan-y-us)]

I believe the duplicate status is wrong. The issue here isn't the actual load cycle count, but our inability to even apply an ugly fix in Hardy, whereas the ugly fixes work in Gutsy.

I have tried EVERY fix that was proposed in the original bug report, and after every resume from suspend or hibernate I'm back to a default APM setting of 128 and frequent clicking (several per minute).

[Valentin Neacsu (valentin.neacsu)]

To set anything other than the default 128 when resuming from suspend/hibernate, set ENABLE_LAPTOP_MODE=true in /etc/default/acpi-support, then apply my ugly-fix from https://bugs.launchpad.net/ubuntu/+source/acpi-support/+bug/89269/comments/18 . After that, anything in /etc/acpi/power.sh under function laptop_mode_enable gets set when on battery, and anything under function laptop_mode_disable gets set when on AC. Hope this helps.

[ethanay (ethan-y-us)]

I have followed your instructions for the ugly fix and hdparm -I /dev/sda still returns an APM of 128 after resume from suspend. I can't find a stray 128 setting anywhere that might be over-riding the ugly fixes, leading me to believe that for some reason anything that I do just isn't registering at all.

Any rate, I think my problem is a bit different, as I can enter hdparm -B 254 /dev/sda manually and it will set the APM value appropriately.

Anyway, this shouldn't be a duplicate of bug #156893 if the person reporting it has an hd* device.