Subject: Re: RAIDframe questions
To: Chris Ross <cross+netbsd@distal.com>
From: Greg Troxel <gdt@ir.bbn.com>
List: current-users
Date: 09/19/2007 16:21:25
Chris Ross <cross+netbsd@distal.com> writes:

>   Okay.  So what are the options for dealing with swap?
>
>   a) Have swap on sd0b or sd1b
> 	This means it will be "bad" if that disk fails, and I want the
> system to keep
> 	running.  It'll not run at all, right?

Yes, if a disk fails, you can crash.

>   b) Have swap on raid0b
> 	This means the above issue is in effect, but swapping will work if one
> 	of the disks fails, and the RAID is running in degraded mode, right?

yes.  This is how I do it.  to me, not using reboot/halt is just
breaking a bad habit - other things might want clean shutdown.

>   There are issues with (b) where it suggests you put a 'b' partition
> on the
> underlying disks.  Is this for the crash-dump code?  Can crash-dumps not
> happen to a raid device?  I guess I need to re-read the aforementioned
> document to check, but any comments from those who've been here
> before would be welcomed.  Thanks!

Yes, the crash dump code is so low level that it doesn't go through
raidframe.  so you can make a real partition that overlays.  But be VERY
VERY CAREFUL.  crash dumps not in the right place have been reported to
cause trouble....