On Wed, 26 Feb 2003 11:28:15 +0000
David Wetzel <dave@turbocat.de> wrote:

> hi folks,
> 
> all my netbsd raid experience is some time ago with scsi (u2w) drives.
> 
> Now, I want to build an server for about 30 (office) clients. The main
> target is data safety and the second is speed. I want to start with:
> 
> 2 120.0GB Seagate Barracuda ATA V drives, NetBSD RAID 1 software raid.
> 
> If the data amount will grow, I will add an el-cheapo PCI-EIDE controller
> card and add another 2 drives. 
> 
> Another option would be to go the "ICP Vortex way" but is that worth the
> money? An Pentium 4 > 2GHz should be able to do it in software fast enough
> I guess.
> 
> I want to be able to boot from any drive, if one fails. (I do not care if
> I have to add a small partition on each drive)
> 
> Any comments/real world expericences? The box will have gigabit lan (on
> Tyan Trinity GC-SL S2707) that should work with the "wm" driver?!
> 

I have three production machines using raidframe. MOst are scsi, but I do
have one machine with two ide drives in a raid 1 setup. Each drive
has 3 partitions. The first partition on each drive holds  a bootable kernel
and a minimal root. The second partions are for swap. The third partitions
are cobbled together using raidframe. After creating the raid set, I created
/, /usr/, etc... on the raid 1 device. If wd1 fails, you can keep going. If
wd0 fails, you need to remove it before you can reboot. This way, you can
have trained monkeys deal with these sorts of things while you're off
skiing. 

I've tried using mutiple dtp hardware scsi raid cards in the past [1.6], but
found that the system would mysteriously lock up under load, or after a few
days. I sure others have had better luck. With hardware raid, you might not
even need the monkeys. So far, Raidframe hasn't let me down and I've
been using it since it's inclusion into netbsd.

-- 
Harry Waddell
Caravan Electronic Publishing
-----------

"I won't eat cereal that doesn't turn my milk purple." - Calvin