Subject: Re: raidframe on a pmax?
To: Aaron J. Grier <agrier@poofy.goof.com>
From: Greg Oster <oster@cs.usask.ca>
List: port-pmax
Date: 07/28/1999 11:50:47
"Aaron J. Grier" writes:
> On Wed, Jul 28, 1999 at 11:18:00PM +1000, Simon Burge wrote:
> 
> > Has anyone had any experience with raidframe on NetBSD/pmax?
> 
> I run a simple RAID1 on my 5000/240.  

Cool.. good to hear.

> It's certainly no speed daemon,
> but it gets the job done.  I've even (accidentally) broken the mirror a
> couple of times and it still keeps chugging along.

:)
 
> > A disk failure on one of the components of my 10x2GB ccd on a 1.3.3
> > system is prompting me to look at alternatives :-(
> 
> :(  With that many drives, you should've seen it coming...  The more
> drives you have, the more chances for failure.
> 
> > The filesystem is mostly read-only, so I was thinking of looking at a
> > raid4 setup with a single hot-spare.  Sound reasonable?
> 
> I'd go RAID5 to balance out the IO.  Even if you're doing mostly read
> only, having an extra spindle available for IO would probably more than
> offset any extra seeking around you have to do on the RAID5 drives with
> data and parity.  If you truly are doing mostly read-only, you might
> even want to think about ditching the hot spare and running a dual
> parity "RAID6" system.  

Umm... RAID6 in RAIDframe is *not* supported in NetBSD at this point.  The 
code is there, but it's a) incomplete, b) buggy, c) completely untested, 
and d) all of the above.  It is on my list of things to get working though..

> If that "hot spare" really is hot, you minus
> well put it to work instead of just hoping that it'll be reliable when
> you actually need it.  Nothing sucks more than having a newly
> reconstructed disk fail.

Well.. maybe nothing except having a 2nd disk die before you get things 
reconstructed :-/

Later...

Greg Oster