Arch Linux

http://mailman.archlinux.org/pipermail/arch-general/2010-February/010987.html

Might want to assign this to Tobias P (tpowa)

I believe that this bug may be caused by the following in rc.sysinit (my version is initscripts 2010.06-2):

# If necessary, find md devices and manually assemble RAID arrays
if [ -f /etc/mdadm.conf -a "$(/bin/grep ^ARRAY /etc/mdadm.conf 2>/dev/null)" ]; then
<------>status "Activating RAID arrays" /sbin/mdadm --assemble --scan
fi

The problem arises when the root filesystem is on a raid array which leads to
having all raid arrays assembled during the initrd boot phase. After which, the
above script snippet runs and tries to assemble an already assembled array set
which then emits a failed exit code from /sbin/mdadm. The status then shows as
FAILED. This is not really a big problem, just annoying.

I can confirm the comment from Tony : this command generates an exit code=2 and hence fails. By the way, this is annoying : I just spend 2 hours looking what was wrong with my RAID array until I stumbled on the above mentioned thread in the forum.

maybe this step should be skipped if udev is running or other check or removed from rc.sysinit...
This step is already done via udev rules, md devices are automatically assembled in incremental way. (-I)

It's not due to the root filesystem (or not just to that, it may have broadened). Unhiding the output shows it trying and complaining about busy items, wrong uuids... but it finally mounts perfectly.

I just happened to have the same problem. I admit it's not a "serious" bug in that it doesn't break something. But I just lost a few unnerving hours trying to figure out why my RAID array "FAILS"... before stumbling over this bug report. In the meantime, a workaround consists in uncommenting the lines related to mdadm in rc.sysinit. But that's a workaround, not a solution.

I can confirm this with an up to date system. RAID works fine though

Is there any way to correct this? I think everyone understands and agrees this has no impact on the system performance or RAID but it just looks bad when a system I dedicate all my time working on as a big red FAIL listed in it. I would think there has to be some way to correct this or patch it...

I'm not sure who activates raid arrays, but on my machine I can comment the section that supposed to do this

# If necessary, find md devices and manually assemble RAID arrays
if [ -f /etc/mdadm.conf -a "$(/bin/grep ^ARRAY /etc/mdadm.conf 2>/dev/null)" ]; then
status "Activating RAID arrays" /sbin/mdadm --assemble --scan
fi

and everything's fine. So the arrays are already active (/ and /boot by initrd, the others by something else, maybe udev?).
It it really needed?

Anyway, if it's needed we can add a check to see if some arrays are not mounted yet before running mdadm --assemble. (I have no idea how to make a patch, here's a simple diff)

140c140,143
< status "Activating RAID arrays" /sbin/mdadm --assemble --scan
---
> if [ $(/bin/grep ^ARRAY /etc/mdadm.conf | /bin/grep -o "md[[:digit:]]" | /usr/bin/sort | /bin/tr -d "\n") != \
> $(/bin/grep active /proc/mdstat | /bin/grep -o "md[[:digit:]]" | /usr/bin/sort | /bin/tr -d "\n") ]; then
> status "Activating RAID arrays" /sbin/mdadm --assemble --scan
> fi

The following patch which is now upstream, let you set USEMDADM=no to fix your issue.

http://projects.archlinux.org/initscripts.git/commit/?id=8b2bfa7bd0073a3a9d389242faf16483c9ec5336