[BlueOnyx:14355] Re: MBR loss on software RAID1

Rickard Osser rickard.osser at bluapp.com
Tue Feb 4 03:35:50 -05 2014


Hi!

Yes, fix_grub was a last-resort solution especially for a combination of
BQ and a 4-disk Tyan solution that sometimes lost it's MBR that I used a
couple of years ago. BO never showed the same problems with the same
hardware. The reason for this is of course the changes in MD that saved
the config to the disks instead of the the raidtab file which all
monitoring script in Sausalito uses. I also wrote a script (lsraid) to
give sausalito the same information as was available on the old systems.

This should all be reformed and updated to modern standards.

Best regards,

Rickard

On mån, 2014-02-03 at 20:27 -0500, Michael Stauber wrote:
> Hi Andreas,
> 
> > I think I've found the cause for the loss of MBR on software RAID1
> > systems. The init script /etc/init.d/mdchk gets executed during boot
> > and on line 227 it writes a syslinux MBR:
> > 
> > system(sprintf("ms-sys -s %s > /dev/stderr", disk));
> 
> Yikes! Great find! Many thanks! Been looking for that issue for ages.
> 
> Yeah, the entire RAID stuff is horribly broken. In 520XR I'll probably
> rip it out entirely and just let a PHP script monitor the RAID status. I
> have that part already working.
> 
> However, the underlying Perl and shell scripts to monitor RAID and to
> initiate a rebuild are simply too old to get them working on EL6 clones.
> I've tried some work arounds, but they'll shoot us in the foot in the
> long haul.
> 

-- 



________________________________________________________________________


Bluapp AB
Rickard Osser
CTO
Solberga Ängsväg 3
125 44 Älvsjö
Sweden




More information about the Blueonyx mailing list