Re: Peculier raidframe problems

[Greg Oster <oster%netbsd.org@localhost>]

On Thu, 18 Dec 2014 20:58:24 +0000
Dave Tyson <dtyson%anduin.org.uk@localhost> wrote:

> I'll got a HP server running NetBSD 6.99.40 (amd64) with 2 pairs of 
> disks each configured as raid 1 mirrors. The first pair of disks has 
> three raid partitions arranged as mirrors on wd0 and wd1.
[snip]
> I noticed that raid2 (/dev/wd0f, /dev/wd1f) had a problem: There were 
> disk errors and /dev/wd0f was marked as failed. I decided to fail 
> /dev/wd0e and /dev/wd0a and replace the disk with an identical one 
> disklabeled as before. I rebuilt the three raids and rebooted.
> 
> The raidsets raid1(/dev/wd0e, /dev/wd1e) and
> raid2(/dev/wd0f, /dev/wd1f) were fine but raid0 had /dev/wd1a was now
> marked as failed... this wasn't the disk that was replaced.
> 
> If I reconstruct /dev/wd1a from /dev/wd0a then everything appears
> fine, but after a reboot /dev/wd1a is marked as failed again. There
> are no disk errors reported
[snip]
> Anyone got a clue how to fix this problem...

What command are you using to reboot?  The component labels should be
already correct once the rebuild completes...   Doing a 'raidctl -I 10
raid0' after the rebuild is done will force the component labels to be
updated... but that shouldn't be needed...

Later...

Greg Oster