Bug#433210: Debian-installer should include /dev/md0, /dev/md1 (plus mdadm, MAKEDEV?)

[debbug . mddev]

Package: debian-installer
Version: 20070308

While trying to perform some rescuing of a Debian Etch x86 system running
with root-on-md (RAID1) (plus some other partitions with md RAID5),
I tried to use the USB Debian Installer. 

This was not as easy as I'd think it would be, because there's no
/dev/md0 accessible from the installer shell. I actually got mdadm
installed by starting the manual partition editor (guess I pulled in
some md installer package) but there were still no /dev/md[0-9] device
nodes around. Also I couldn't find any MAKEDEV script so creating them
were not trivial. Finally I managed to put the correct md device files
on a mountable file system and "cp -a" them into the running installer
shell.

So in summary, I'm submitting this bug report in the hopes that rescuing
md systems can be made easier with the next version of the debian 
installer.

PS: I also got a whole lot of scary messages in 'dmesg' when the
debian-installer automatically tried to mount every partition to look
for the .ISO file (it tried to mount the raw devices, which are part of
undetected MD arrays). Hopefully the mounting done by the .ISO detector
doesn't un-sync the arrays (does that make sense?)



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Bug#433210: Debian-installer should include /dev/md0, /dev/md1 (plus mdadm, MAKEDEV?)

[debbug . mddev]

On Sun, Jul 15, 2007 at 17:50:24 +0200, Frans Pop wrote:
> On Sunday 15 July 2007 15:46, [EMAIL PROTECTED] wrote:
> > While trying to perform some rescuing of a Debian Etch x86 system
> > running with root-on-md (RAID1) (plus some other partitions with md
> > RAID5), I tried to use the USB Debian Installer.
> >
> > This was not as easy as I'd think it would be, because there's no
> > /dev/md0 accessible from the installer shell.
> 
> You seem to be somewhat confused about how /dev/md* devices are created. 
> They are never created manually, but are instead created automatically by 
> the kernel when RAID devices are activated.
> 
> What you need to do to rescue a software RAID system is:
> - manually load the needed kernel modules ('modprobe raid1')
> - run mdadm to assemble existing arrays:
>   # mdadm --examine --scan --config=partitions > /tmp/mdadm.conf
>   # mdadm --assemble --scan --run --config=/tmp/mdadm.conf --auto=yes
> 
> Closing your report as this is not a bug, although we do agree that better 
> support for software RAID in rescue mode would be nice, but that is a 
> known issue.

Thanks for the scan hint, I wasn't aware of that.

Unfortunately I don't have access to the faulty system right now so I
can't re-check, but I was pretty sure that I tried something like 
"modprobe md; mdadm --assemble --no-degraded /dev/md0 /dev/sda2 /dev/sdb2 
/dev/sdc2" 
and it complained about /dev/md0 not being there, then, after I got that
device node in place, the same command worked. This is from memory, 
so I might be wrong though. 

Anyway, sounds like you're aware of the issues (more than me :)) so 
I agree there's no need to keep the bug open.

Thanks again!



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]