[Nottingham] Big biiig problem regarding IDE-RAID

Robert Davies nottingham at mailman.lug.org.uk
Sat Jul 19 18:46:01 2003

On Tuesday 15 Jul 2003 15:56, Graeme Fowler wrote:
> On Tue, 15 Jul 2003, Robert Davies wrote:
> > There's some nice-ish IDE RAID enclosures with SCSI interfaces, with
> > serial ATA coming hot swap IDE is going to be made simpler, so costs
> > should fall yet more.
> Indeed they will, and I look forward to that day with glee :)
> > What you're thinking of is very very doable with software RAID.
> Almost certainly. However, one very major difference between straight
> kernel-side software RAID and this specific chipset/driver combo seems to
> be that there is much more of the control being done on the chipset, and
> the driver is simply reading results from and sending instructions to it.
> It's currently running a rebuild after we forcibly removed a drive during
> Upon rebooting, the BIOS gave us the option to either rebuild from within
> the BIOS or to continue booting. Continuation resulted in the system
> rebuilding "backgrounded", with progress reported in /proc/scsi/FastTrack/0
> Notably, it's not reporting scads of activity through vmstat or top, so it
> appears that the driver is simply saying "invoke rebuild from drive 0 to
> drive 1" to the chipset, and not troubling itself with the data. Which is

I just synced up a software RAID mirror and took care to note vmstat 5 output.  
The CPU% was lowish, +90% idle time but 6-7% System usage, a x5 interrupt 
rate, and roughly 10x context switch rate than normal.

With SCSI disk the interrupt rate would be a bit lower, but as system very 
usable during this sync operation, remaining responsive and you don't do that 
very often, who cares?  That's not always been the case with some RAID Tower 
features, where I've found reconfiguring online causes the thing to run like 
molasses, and would be unacceptable for a production server.