[CentOS] sleepy disk drives

Aleksandar Milivojevic

alex at milivojevic.org
Wed Feb 20 17:26:44 UTC 2008


Hi,

I'm having a grief with Seagate FreeAgent Pro drive connected to
CentOS 4 box via USB.  The problem is that drive is factory configured
to take a nap (spin down) after some 15-20 minutes of inactivity, and
there doesn't seem to be a way to disable it.  Since it's USB
connected, hdparm can't be used on it.  Getting the drive out of the
case is not an option either, since it would probably void the
warranty (and there's still 4.5 years of it left).  There's eSATA
connector on the drive too, but I don't have anything on the PC side
to connect it to.  There's also FireWire connector, but somebody at
big red company decided users don't need FireWire (and I'd probably
have same problem if the drive was connected via FireWire too).

The problem with the drive spinning down is that Linux doesn't really
like when that happens.  It reports that drive is not responding,
spits bunch of I/O errors for the device in the logs, and refuses any
operation on the file system(s) that are on that drive.  In contrast,
Windows, Mac OS X and some other boxes I tried to connect the drive to
have no problem with it (drive simply spins up when accessed).  There
was a patch in relatively recent Linux kernels that adds allow_restart
option (accessible from /sys), which apparently should solve the
problem for Linux (described here:
http://www.nslu2-linux.org/wiki/FAQ/DealWithAutoSpinDownOnSeagateFreeAgent).
 When allow_restart is set to 1 (default is 0), Linux will do what all
other operating systems do by default, spin up the drive on access.
However, kernel included with CentOS 4 is not recent enough, and
doesn't have that option.

Anyhow, my question is (after probably describing things a bit too
much), does the kernel included with CentOS 5 supports allow_restart?
Could somebody (anybody) running CentOS 5 could check if either of
these paths exists on his/her system:

/sys/block/sda/device/allow_restart

or

/sys/class/scsi_disk/0:0:0:0/allow_restart

If they do exist, I'll be probably upgrading my box to CentOS 5
pronto...  If they don't, well, might still upgrade but it won't solve
the problem I'm having...

Thanks,
Alex



More information about the CentOS mailing list