On Sun, 5 Nov 2006 01:05:08 -0500, Kevin Brunelle <kruptos@mlinux.org> wrote:
> On Sunday 05 November 2006 00:17, Malcolm Herbert wrote:
>> How does this behave in cases where we are using RAIDFrame? I'm using it
>> for my root disk and booting from it directly ... if the kernel needs to
>> create such a dump, how will this interact with RAIDFrame when it tries
>> to dump to /dev/raid0b as advertised?
> 
> You should not do that.
> 
> Here's a good article explaining what to do
> http://users.piuha.net/martti/comp/raid/raid.html but I will try and cover
> the relevant parts.  Note, I don't currently have this setup myself but I
> am
> looking into it for someone.
> 
> ---
> /dev/raid0a     /               ffs     rw      1       1
> /dev/raid0b     none            swap    sw      0       0
> /dev/raid0e     /home           ffs     rw      1       1
> /dev/wd0b       none            swap    dp      0       0
> ---
> 
> In this case, /dev/wd0b is aligned with the part of /dev/raid0b that is on
> wd0... you'll have to read the article to get a full understanding.
> Basically, this way the kernel knows the actual disk areas where to dump
> the
> information.  You need to use swapoff to avoid parity errors and such.
> 
> In practice I am not sure how well all this works or what effect it has on
> swapping.  But if you're concerned about this and you are unlikely to ever
> use a dump... just turn off dumps entirely.
> 
> Add the following line to /etc/sysctl.conf
> kern.dump_on_panic=0
> 
> -Kevin
> 
> I hope you're less confused than I am after writing that.  Basically,
> don't
> dump to raid0b... either use the solution from that link or turn dumping
> off.

There's also some info about this in the official NetBSD guide. I run my root filesystem on RAIDFrame, and found the NetBSD guide to be very helpful.

Info here:  http://www.netbsd.org/guide/en/chap-rf.html#chap-rf-setup-kerneldump

Cheers,
Chris.