First things first. I'd say this is not an issue that should stay this
long in 'Undecided'-state. Having to manually resync RAID-arrays after
each reboot is a real issue to a desktop user (of course one could
always question why on earth would this desktop user use RAID but that
is besides the point).

This bug really seems to be related in the bug described in
http://lkml.org/lkml/2006/10/6/233 or at least the behaviour seems to be
almost exactly alike.

I've added some extra output commands to /etc/init.d/mdadm-raid (see
below for diff against original file). This reveals that it is the root
device, /dev/md0, that is failing each and every time (well, to be
honest, I only tested this three times after jumping into conclusion by
bold extrapolation).

When init runs "/etc/init.d/mdadm-raid stop", as one of the last actions
during the shutdown, for some odd reason /dev/md0 is still on active
state but every other array is already in stopped state. Now this is as
it should be, since we don't want those arrays to be busy while stopping
them. Now what comes to my mind is this is not exactly a bug in mdadm
but instead a bug in kernel itself (as that lkml thread hints).

234,240d233
< 
<         # show /proc/mdstat after each issued stop command
<         cat /proc/mdstat
< 
<         # show what exactly has been run
<         echo "C: \`$line'"
<

-- 
mdadm fails to stop RAID on shutdown
https://bugs.launchpad.net/bugs/111398
You received this bug notification because you are a member of Ubuntu
Bugs, which is the bug contact for Ubuntu.

-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to