[BlueOnyx:11168] Re: MDADM / FDISK problem

[Michael Stauber]

Hi Richard,

> I've got a bit of a problem with disks.  It was reporting a disk
> failure, but when I replaced the faulty disk last week it didn't do the
> RAID rebuild.

I had a similar problem recently and this appears like the cause could
be the same here:

During install off the CD the disks are partitioned with Anaconda's Disk
Druid. Which is fine.

During RAID rebuilds fdisk is used to try to create the same partition
layout on the replacement disk which exists already on the good disk of
the RAID1 array.

However: Disk Druid and fdisk partition slightly differently and
typically fdisk has a hell of a lot of problems creating the very same
partition size.

Unless you help it along with sfdisk, which allows us to copy the
existing partition table from one disk to another disk.

I went through a metric ton of emails, after remembering that we had the
same problem in 2006 and it had been fixed in mdchk. But somehow it
apparently fell out of the BlueOnyx code again, because I don't see it
in our current mdchk.

Please run this command:

/usr/bin/perl -pi -e "s/sfdisk -L/sfdisk --force -L/"  /etc/init.d/mdchk

Then simply run this:

/etc/init.d/mdchk

Finally check with "cat /proc/mdstat" if your RAID is rebuilding again.

Let me know if that works and I'll publish a fixed mdchk init script via
YUM.

-- 
With best regards

Michael Stauber