Re: [smartmontools-support] inactive SATA drives won't stay in standby or sleep, PATA models did. (fwd)

[Tejun H. <ht...@gm...>]

Hello,

Sorry about late response.  I've been traveling for more than a month.

Bruce Allen wrote:
> Hi Tejun,
> 
> FYI.  Feel free to ignore this, or you can respond directly to user/list
> if desired.

cc'ing smartmontools-support and the original reporter.

> ---------- Forwarded message ----------
> Date: Sat, 13 Sep 2008 17:42:29 -0700
> From: Linda Walsh <sma...@tl...>
> To: sma...@li...
> Subject: [smartmontools-support] inactive SATA drives won't stay in
> standby or
>     sleep, PATA models did.
> 
> I'm having problems with disks staying "asleep" or "suspended" (spun-down).
> 
> I've been trying to monitor the temperatures on the disks to help
> note cooling problems.  I use the "smartctl -n standby -A <device>"
> command to spew out the attributes and look for "Current Drive
> Temperature" or Attribute#194 (or I look for and print STANDBY|SLEEP
> if that is found).
>
> I can force the drive to standby or sleep using the -y or -Y command
> work).  I've also made it go to sleep by setting the drive timeout
> to 5 seconds (-S 1).  But if I run my "drive_temp" command a few
> times the drive will go from 'STANDBY' back to running in "fairly
> short" order: usually about 30 seconds.
> 
> When I'm running the monitoring script to poll every 10 seconds, I
> can see when the script is going to return the temp -- because if it
> is in STANDBY, I get maybe 2 reads, then on the 3rd, it pauses when
> I issue the smartctl command and waits for it to spin-up and then
> gives me the temperature.

Bruce, is there a smartctl option to tell us what's going on?

> It's on two different drives that I have observed this -- both
> Seagate's, one a 750G, the other a 1000G.
> 
> I tried running the 'short' tests on each (as someone else had a
> similar problem that seemed to be fixed after running the short
> drive health tests).  The drives do claim to be in "standby", but
> keep spinning back-up.

Hmm... sounds like coincidence to me.

> Both drives that are doing this are "backup" drives.  I.e. all they
> store are daily system backups -- so they should only turn on in
> early "AM" to receive the backups, but then should time-out.  They
> both *did* timeout (goto sleep) regularly, when both drives were
> PATA Seagates.  But now, I can't keep them asleep.
> 
> It's only my SATA drives that 'should' be going to sleep now.  I
> have the two SATA drives on a Promise SATA-300 TX4 (4 internal Sata
> ports).  My "active" (only 2 other) drives are on different
> controllers a SCSI and a PATA port.  They don't sleep or spin-down.
> 
> I noticed the following when I tried to use the sleep command on the
> system console:
> 
> 17:23:26 Ish kernel: ata2.00: exception Emask 0x0 SAct 0x0 SErr 0x0
> action 0x2
> 17:23:26 Ish kernel: ata2.00: waking up from sleep
> 
> I don't see the console messages when trying to wake it up from standby.
> But should I be getting kernel error messages on a wakeup from drive sleep?
> (kern=2.6.25.12, vanilla)

Any command issued to a sleeping drive triggers wake up action as
otherwise it will just gonna timeout, so that's libata telling you
that it's waking up the drive to process whatever pending command.
Hmmm... it seems there needs to be a way to export that the drive is
sleeping to userland.

Thanks.

-- 
tejun