3

I set up a home server about a year ago, using Ubuntu server (10.04 LTS at the moment), four disks in RAID 5 for storage (WD Green 1.5 TB) and a laptop drive for the OS.

Today the output of smartctl, a command line utility for checking the SMART attributes of a hard drive, tells me that the primary OS drive has had no less than 186,000 start-stop cycles in 325 days and may be nearing the end of its lifespan.

The smartctl output is in "normalized values", in this case a number between 200 and 000, where 200 is "brand new" and 000 means "worn out". My disk gets 001.

So I wonder what happened: 186k start/stop cycles in 7820 hours is about one start/stop per 2.5 minutes around the clock. This seems somewhat excessive for a computer that sees actual use once or twice per day. (The RAID disks are normal, averaging to one start/stop per day, as expected.)

Does anyone have similar experiences, or pointers to what might be the issue here?

Specifically I'd like to know

  • Why the massive start/stop count? Do I have some sort of configuration issue? Could there be a background service that is causing trouble?
  • Could having a laptop disk as the OS drive be part of the problem? Can anyone confirm or deny this?

Here is the /etc/hdparm.conf configuration

/dev/sda {
  apm = 127
  spindown_time = 120
}

and the most relevant parts of smartctl --attributes /dev/sda:

smartctl version 5.38 [x86_64-unknown-linux-gnu] Copyright (C) 2002-8 Bruce Allen

=== START OF READ SMART DATA SECTION ===
SMART Attributes Data Structure revision number: 16
Vendor Specific SMART Attributes with Thresholds:
ID# ATTRIBUTE_NAME          FLAG     VALUE WORST THRESH TYPE      UPDATED  WHEN_FAILED RAW_VALUE
  1 Raw_Read_Error_Rate     0x002f   200   200   051    Pre-fail  Always       -       0
  4 Start_Stop_Count        0x0032   001   001   000    Old_age   Always       -       185875
  9 Power_On_Hours          0x0032   090   090   000    Old_age   Always       -       7820
 12 Power_Cycle_Count       0x0032   100   100   000    Old_age   Always       -       109
193 Load_Cycle_Count        0x0032   118   118   000    Old_age   Always       -       246833
194 Temperature_Celsius     0x0022   107   098   000    Old_age   Always       -       36

As I generally prefer my drives to last more than a year, any advice is appreciated.

Update

Apparently the "apm = 127" setting in hdparm.com was the problem. Commenting out the settings I get the default, 254, and the disk never spins down at all.

That's not quite what I was looking for either, I'll have to see if I can find a middle ground somewhere. Still, the problem from the title of this post is solved. Thanks for your help.

Some more detail for the next person with similar problems:

apm is Advanced Power Management, a value 1-255. Higher values mean "more performance", lower values "more power saving", 255 is "disabled".

I had picked 127 as the "highest performance that still allows disk spindown" according to hdparm man pages, as I wanted the disk to go to sleep when the server was not in use.

What it got me was the manufacturer's 20-second default spindown time for this particular drive (a WD Scorpio Blue), a fair enough default for a laptop running on batteries.

With the OS writing to disk all the time (system logs and such, whether or not the computer is in actual use), the disk would barely fall asleep before being awoken again, and I got the start/stop every 20 seconds behavior. My attempt at increasing the spindown time (I had set it to 10 minutes) was apparently ignored by the drive.

At some point I had installed laptop-mode, which caches disk reads/writes in memory, so the OS was only supposed to write to disk every couple of hours.

The primary problem in this case was that laptop-mode stopped working after an upgrade - it is still listed as a service to start at bootup, but it no longer starts. And I had more or less forgotten about it and didn't think of checking.

At least I know where to look now, thanks again for your input.

1

3 Answers 3

1

Some things to check for:

  • Is the problem occurring now? (sample the drive, wait a day, sample again, and see if it increases noticeably (say once for every 2.5 minutes in a day)
  • Is the problem occurring for all of the disks, or just one?
  • What is the power configuration for the computer? Power saving, or no? Spin down the disks, or no? Check hdparm -B and hdparm -S (and read the man page for information on how to interpret the data)

If the problem isn't happening now, I do recall a bug that was reported related to hard disks spinning down and up repeatedly in Ubuntu, but it may have been a while ago. You might investigate that, see if maybe it was fixed in an upgrade.

If the problem is only for one disk, you have to ask what is special about that disk.

If the settings above don't match your needs, they may be related to the problem, or even the culprit.

1
  • It's occurring now, yes. It seems to be getting worse, it had 400 restarts in about two hours, or once per 20 seconds. To the best of my knowledge that's beyond configuration issues, it must be something seriously wrong somewhere... Still thanks, I'll try out some restarts with different settings. Jun 18, 2010 at 5:30
3

Oh yeah, that problem...

This should work.

2
2

What is the manufacturer of the laptop drive? Is it also WD like the ones in the RAID? The Start_Stop_Count is defined as "Number of start/stop cycles of spindle" for the drives that I saw in the smartmontools faq. Unfortunately I don't have a large number of drives to compare to this one. My laptop has a western digital drive and I am also running ubuntu. It is several years old and only has about 300 of the same attribute. There are many things that can be going on here.

  1. Apparently different manufacturers report their data differently from one another (even across brands of the same manufacturer). So the number might mean something other than what the label suggests.

  2. Laptop drives might have power saving logic built in which might involve stopping the spindle from spinning when idle.

  3. The sensor that reports the attribute to smart might be broken.

  4. The most likely cause is that Ubuntu is set to spin down the drive when idle. You should check out this post.

1
  • Yes, it's a WD Scorpio Blue. Thanks for the link, I'll check it out. Jun 18, 2010 at 5:26

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .