spinning down drives can cause nightmares.  I've run across a frighteningly
large number of drives which wouldn't spin back up until they were power
cycled.  I haven't been keeping a good catalog so i'm not sure which cases
are age-related and which could be firmware problems.

x86 systems have the additional problem that they need to have their SCSI
bios configured to spin the drives up when the system POSTs or you can end
up hanging the system in the bios forever.  Unfortunately many of them don't
seem smart enough to deal with a "not spinning" check condition in any
intelligent way other than retrying.  Presumably non-x86 systems will have
less problems, but it depends on what the firmware's going to do.


ideally flushing won't be too expensive if the disk's not actually caching
any write-data.  Personally i don't trust hard drives even after i've turned
off write caching, so sending the flushes is probably the safe thing to do.

-p


-----Original Message-----
From: Jason Thorpe [mailto:thorpej@nas.nasa.gov]
Sent: Tuesday, June 09, 1998 2:04 PM
To: Matthew Jacob
Cc: tech-kern@netbsd.org
Subject: Re: test of new powerdown facility 


On Tue, 9 Jun 1998 13:55:25 -0700 
 Matthew Jacob <mjacob@feral.com> wrote:

 > I still want to know why we would do this if disks aren't set
 > to cache data? 

I think the question is "Can it hurt?"  If the drive isn't set to writeback,
then there won't be any remaining blocks to sync, which is not an error.

I think I'd prefer to err on the safe side :-)

I think Gordon's "spin down" idea is good, too... once the sync operation
is complete, maybe spinning down the disk isn't such a bad idea, too...

Jason R. Thorpe                                       thorpej@nas.nasa.gov
NASA Ames Research Center                            Home: +1 408 866 1912
NAS: M/S 258-5                                       Work: +1 650 604 0935
Moffett Field, CA 94035                             Pager: +1 650 428 6939