RAIDframe as root and kernel dumps

Hello,
Iam just setting up my home server with a mirrored RAIDframe
root.

Now I'm not a NetBSD developer and I wouldn't know how to
debug a
kernel crash dump, even if it came after me with fork and
knife, so is
there any reason why I should bother setting up kernel
dumps, as it is
mentioned in the guide [1]?

Is there any other reason or is it save to just use swap
inside of the
raid and not having space for crash dumps?

Cheers
Felix

[1]http://www.netbsd.org/docs/guide/en/cha
p-rf.html#chap-rf-setup-kerneldump

P.S.: Please CC me, as I have to use crappy webmail at the
moment.

Sorry I was in a hurry:

2007/7/18, Felix 'buebo' Kakrow <buebobuebo.de>:
> Hello,
> Iam just setting up my home server with a mirrored
RAIDframe root.

Which is a NetBSD 3.1 on i386 system.

Thanks in advance!
Felix

"Felix 'buebo' Kakrow" <buebobuebo.de> writes:

> Iam just setting up my home server with a mirrored
RAIDframe root.
>
> Now I'm not a NetBSD developer and I wouldn't know how
to debug a
> kernel crash dump, even if it came after me with fork
and knife, so is
> there any reason why I should bother setting up kernel
dumps, as it is
> mentioned in the guide [1]?

The only reason is that if you have trouble, you can get a
backtrace and
post it, and that may be helpful if you have a problem that
others do
not have.

> Is there any other reason or is it save to just use
swap inside of the
> raid and not having space for crash dumps?

As the guide says, setting up kernel dumps to overlap with
raid swap is
dangerous.  So you have several choices:

1) don't make swap big enough for dumps, and never have
dumps

2) make swap big enough, and don't set them up, letting you
do it later

3) make swap big enough, and set up dumps

4) put swap not on raid, and make it big enough, and maybe
dump on it.
This setup means that a disk error in swap will take your
system down,
but if your concern is saving the data and not availability,
that may be
ok.  My raid systems are all servers (4) and i have
everything in raid
including swap because I want them to keep running if
there's a disk
failure.

Given what you said, I'd go for 2 or 4.

On Thu, 19 Jul 2007, Greg Troxel wrote:
> The only reason is that if you have trouble, you can
get a backtrace and
> post it, and that may be helpful if you have a problem
that others do
> not have.
>
>> Is there any other reason or is it save to just use
swap inside of the
>> raid and not having space for crash dumps?
>
> As the guide says, setting up kernel dumps to overlap
with raid swap is
> dangerous.  So you have several choices:
>
> 1) don't make swap big enough for dumps, and never have
dumps
>
> 2) make swap big enough, and don't set them up, letting
you do it later

What I've tried to do is to create a shadow disklabel entry
with a start 
offset of 64 more than the RAID partition for swap and a
size of 64 less 
with type swap (IYSWIM). I then set this (on the first disc)
as the dump 
device. However, it complains that dumping to dev 18,8
offset xxxx is not 
possible. I'm not sure why, the same trick works for
filesystem if the 
shadow entry is 4.2BSD.

-- 
Stephen