Re: RAIDframe and softupdates, boot time after a crash

On Sun, 19 Jan 2003, W.E.B. Schrott wrote:

> There was some hints in a mail to misc@ lately (Alex
> Cichowski: Re: RAIDFrame Issues on 3.2, 14.01.2003
> 13:58), but has anyone a raid 1 setup which checks on
> startup (in /etc/rc) whether the server crashed and
> the parity is dirty and then marks one as degraded and
> goes on booting. Then when up and running (with a
> broken mirror r-enable the mirror by writing parity to
> the degraded drive?)
>
> Can anyone paste their modified start scripts and
> config or sketch the appropriate steps needed?

What I was thinking of when I wrote the message you mention was doing something like this in /etc/rc before "raidctl -P all":

# Check raid0 parity state
if ! raidctl -p raid0; then

	# If no component has already failed, fail one to avoid
	# a parity rewrite
	if ! raidctl -s raid0 | grep ': failed' >/dev/null; then
		raidctl -f /dev/somecomponent raid0
	fi

You could either schedule a cron job to start the reconstruction, or perhaps begin reconstruction immediately in the background by adding "raidctl -R /dev/somecomponent raid0 &" after the "raidctl -f" line above.

I haven't actually tried anything like this yet, though.

Alex Received on Mon Jan 20 14:03:53 2003