[clug] Ubuntu problem with mdadm / kernel?

Edward C. Lang edlang at edlang.org
Wed Sep 5 18:44:38 MDT 2012

I've rebooted the server in question many times in the last few months
since I built it late last year, and the /etc/mdadm/mdadm.conf file was
last modified in December. I'll have a look at the differences in the
output tonight...

I'm not aware of any hard or soft errors in any of the RAID devices. I
wondered if the issue was that the affected RAID1 device was essentially a
second tier - none of the other devices (RAID1 /boot with four submirrors,
RAID1 / with four submirrors, and the two RAID1s that are under the RAID0)
had issues. Perhaps mdadm is now doing fewer passes when scanning than it
did in the past?

(I really don't like how easily Ubuntu will pause the boot process if a
filesystem is missing. Is there some way to set a timeout for that?)

On Thu, Sep 06, 2012 at 10:28:44AM +1000, Steve McInerney wrote:
> I've had issues in the past, not necessarily ubuntu specific, where
> the mdadm.conf becomes out of sync to reality; usually to changes
> I've done, but not updated.
> sudo mdadm --examine --scan
> and diff to /etc/mdadm/mdadm.conf may be worthwhile.
> the 11.10 -> 12.04 beta cycle was quite nasty for me. Ended up
> calling it quits, doing a full wipe and reinstall from the 12.04
> beta of the then moment. Some related to mdadm going lala on me, if
> memory serves.
> Had happily upgraded thru each beta cycle since 8.10, so it was
> crushing to do a fresh & full re-install :-)
> You *may* have fallen afoul of:
> https://bugs.launchpad.net/ubuntu/+source/mdadm/+bug/946758
> whereby the mdadm monitoring aborts; so you lose sight of monitoring
> not happening. Which then cascaded to what you saw?
> ??
> Cheers!
> - Steve
> On 06/09/12 10:06, Edward C. Lang wrote:
> >Hi,
> >
> >Is anyone aware of a recent (last three weeks?) issue with software raid
> >devices on Ubuntu? I was away for a few weeks, came home on Sunday and did
> >a dist-upgrade on a headless fileserver. When I rebooted it, for some
> >reason one of the metadevices wasn't found - a 6TB RAID0 device on top of
> >two 3TB RAID1 devices. There are five metadevices in total and the RAID0
> >one was the only one which wasn't automatically detected.
> >
> >When I finally got around to determining why the server wasn't coming back
> >up, I was able to manually run mdadm --assemble --scan, which found the
> >missing device.  At that point I could see the LVM volume group and the
> >server was able to continue booting normally.
> >
> >Perhaps the issue isn't the kernel or mdadm? Could some other package be
> >causing the problem? There wasn't anything obvious on the Ubuntu
> >bugtracker for mdadm or the kernel, and the #ubuntu channel on freenode
> >was fairly useless.
> >
> >Regards,
> >
> >Edward.
> >



More information about the linux mailing list