Lars NoodC)n wrote:

quoted text
> I'm looking to set up a test unit this summer to try RAID 1+0 or 3+0.

> The goal is to be able to stream data quickly but tolerate loss of at

> least one disk at a time.

>

> It looks like there are three options supporting at least RAID level 0

> and 1, can/should any of these be stacked yet?

>

>  softraid

>   http://www.openbsd.org/cgi-bin/man.cgi?query=softraid

>

>  raidframe

>   http://www.openbsd.org/cgi-bin/man.cgi?query=raid

>

>  ccd

>   http://www.openbsd.org/cgi-bin/man.cgi?query=ccd

>

> It looks like one disk (or CF) is needed for a base OpenBSD system, and

> the other 4+ for RAID.  Is that the best way to ensure unattended booting?

>

> Also, what kind of data would you want back on the tests?  It's early

> enough that I can work it into the plans.

>

> Regards,

> -Lars

>

>

Hey Lars,

I'm using raidframe with 1+0.  From my reading of the raidctl(8) man

page it looks like you can compose raid sets from any valid devices,

including other properly configured raidframe devices.  If you're

interested I'll send you my /etc/raid*.conf files.  I'm sure you know

this but for posterity, read the raidctl(8) man page closely.

Everything is in there.  Nevertheless, it took me two passes to figure

out how to get auto configuration working correctly.

As for test results, I'd be interested in throughput performance when a

drive is taken out of the pool (failed) and throughput performance while

the array is rebuilding once a spare has been brought online.
Cheers,
Aaron