On Fri, Sep 20, 2002 at 11:40:43PM +0200, Wojciech Puchar wrote:
> > WARNING: raid0: total sector size in disklabel (232823808) != the size of
> > raid (232824704)
> >
> > 232823808 blocks is the total size of the RAID disk in the disklabel.
> > 232824704 blocks is a weird number, a bit (112 blocks) smaller than the
> > number of blocks I have allocated for the RAID system on the disks.
> 
> it's normal than raid device size is bit smaller than size of component (i
> think it's RAID1 from your mail) - few sectors are used for raid headers,
> parity state info etc etc.
> 
> but it's strange that raid disklabel got bad data. how did you made
> disklabel?
> 
> i'm  always doing disklabel -I -i raid0 on fresh raid. and it works fine.

I created the label manually and then used disklabel -r -w raid0 foo

I guess I should have obeyed the man page..

Do you know what values I should use in /etc/raid0.conf for fast disks?
I'm using

START layout
# sectPerSU SUsPerParityUnit SUsPerReconUnit RAID_level_1
128 1 1 1

START queue
fifo 100

for a RAID1 system, but extracting tar balls feels very slow.  My disks are
120 GB WDC 7200 rpm disks with 8 MB caches and when I tested the read speed
with dd (from raw device), I got 49 MB/s from both disks _simultaneously_.
I think the disks are fast, but RAID is killing the performance.  My CPU is
a 700 MHz Athlon in this case.

I'll Cc: to the list..

  -jm