Subject: Automatic dump handling for swap-on-raid
To: None <tech-userlevel@NetBSD.org>
From: Martin Husemann <martin@duskware.de>
List: tech-userlevel
Date: 09/21/2007 10:05:32
One point that comes up every time someone tries to setup a new system using
raid 1 as root is the handling of swap-on-raid vs. dumps.

The guide describes the setup magic needed to do this, but I wonder if
we should make the process more straight forward with modern means available
now.

How about we add a special rc.conf variable like swap_on_raid=raid0b that
causes some rc.d scripts to do the proper magic:

 - check the physical position of the swap-on-raid partition
 - do the necessary calculations and create a wedge that maps this
   to the first underlying disk device
 - use swapctl -D to define this wedge as dump device

This, of course, would need to run before savecore.

Sounds doable to me, and only loses the ability to dump before reaching this
rc.d script, so a kernel developer would need to still do it the old way
(or not use swap on raid on test systems).

Thoughts?

Martin