Re: Adding disks with raid to existing raid system.

On Sat, Mar 20, 2010 at 5:30 PM, Simon Matthews
[off-list ref] wrote:

On Mon, Mar 15, 2010 at 9:59 PM, Michael Evans [off-list ref] wrote:

quoted

On Mon, Mar 15, 2010 at 7:53 PM, Simon Matthews
[off-list ref] wrote:

quoted

On Mon, Mar 15, 2010 at 5:36 PM, Michael Evans [off-list ref] wrote:

quoted

On Sun, Mar 14, 2010 at 10:23 PM, Simon Matthews
[off-list ref] wrote:

quoted

I have just built a system and have it booting off a software raid
partition. The raid sets use devices /dev/md0, /dev/hd1, /dev/md2,
/dev/md3.

I now need to transfer some additional disks to this system. These
disks are presently in another system where they host a number of raid
sets, currently also /dev/md0 - /dev/md4.

I need to ensure that the data on the raid set that I am adding to the
system is not lost. However, clearly, I can't have the raid sets on
these disks come up as /dev/md0-md4. How do I ensure this and have
these raid sets come up on /dev/md5 and higher?

Simon
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Either use an mdadm.conf to specify the mapping of UUID to md device
(which will over-ride any auto-detected requests), or use the
home-host fallback.  Obviously the administrator specifying how they'd
prefer mdadm to assemble the drives is preferable.

I'm not aware of the "home-host fallback" can you give me some pointers on this?

quoted

You will probably want to regenerate your initrd; if you are using
auto-assembly on root without an initrd, I highly suggest upgrading to
use an initrd/initramfs.  You might find this one easy to customize
for your needs if your distribution lacks one or you dislike the one
it generates: http://sourceforge.net/projects/aeuio/

Fortunately Gentoo includes mkinitrd so I can try this if other
methods don't work reliably.

Simon

quoted

man mdadm
/host

--homehost=

This will override any HOMEHOST setting in the config file and
provides the identity of the host which should be considered the home
for any arrays.

I tried creating an initrd, without any great success so far (would
not boot). But while I work on that, I had another thought.

If I make the partition types for the raid components type 83 (ext3),
the kernel should not (I think) recognize and start these arrays, but
they will be stated later in the boot process by the userspace  tools
as long as there are appropriate entries in /etc/mdadm.conf.

Does this sound like a workable solution?

How can I tell what HOMEHOST parameter is on the raid sets? I don't see it with
mdadm --examine /dev/sdXX
or
mdadm --detail /dev/mdX

Simon

It depends upon your system having a valid /etc/mdadm.conf file with
the proper arrays in it.  If it was unable to mount your root device
aeuio should drop you in to an emergency shell.  Not all the commands
you typically use will work there as it uses either the very minimal
set provided by klibc or busybox.

You might try invoking /bin/ash to get a more complete environment (in
case you had both klibc and busybox klibc is preferred for /bin/sh .
It provides the absolute-core set of commands and is smaller than
busybox, but support for modules currently requires modprobe (it's
easier to use than a dedicated shell-wrapper for insmod) which might
pull in busybox if you have it as well).

You should also look at the messages right above the emergency shell.
In case you /still/ can't start it, the initramfs includes a script
under /etc/init.d/mdadm-probe-all ; It's fairly aggressive, and boils
down to three steps.

1) Make sure mdadm exists
2) (move any existing mdadm.conf file to mdadm.conf~ and then)
Run mdadm --examine --scan /dev/[sh]d* /dev/mapper/* > /etc/mdadm.conf
3) Run mdamd --assemble --scan --no-degraded

The final statement is important, as there is no safety net for
degraded devices.  At that point if your array is still stuck it's an
Administrative Decision point.  You can drop the --no-degraded part if
you absolutely have to get inside, or run any mdadm commands you need
to try and repair things in single-user initramfs mode.  You could
also pop in a recovery disc or USB drive and boot a recovery
environment.
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html