Subject: raidframe hiccups in 1.4S
To: None <port-alpha@netbsd.org>
From: Aaron J. Grier <agrier@poofygoof.com>
List: port-alpha
Date: 02/15/2000 22:45:17
sources supped 2000-02-11  (+/- one day)

savecore: reboot after panic: vwakeup: neg numoutput

I finally knuckled down and wrote a script to help me figure out the
ideal sectors per stripe unit which does this:  (this crash came up
while trying to mount a newly configured raid.)

changes /etc/raid0.conf
runs raidctl -c
changes the disklabel and formats if necessary
mounts the raid, runs bonnie, unmounts the raid
runs raidctl -u

I'm compiling a new kernel (and raid configuration tools) as I type and
will provide any interested parties in my hackish bourne shell script
which performs the above...

I'm also interested in input for configuring raids...  (ISTR Greg Oster
having this discussion on -current a while back, but can't seem to find
the magic rules for RAIDFrame configuration.)  I would normally think
that four disks could get more throughput than one, but I can't seem to
get more than 65% of the block output or 91% of the block input for the
raid compared to a single disk in the raid.  Seeks are another story --
I'm getting 123% more seeks... but the block input disturbs me.

-- 
  Aaron J. Grier | "Not your ordinary poofy goof." | agrier@poofygoof.com
  "Time Correct function allows automatically correcting slight variation
   of your key touching manner."  --  Roland MSQ-700 manual