FC3 T1 Raid Installs

[Alexandre Oliva]

On Aug  6, 2004, Russell Coker <russell at coker.com.au> wrote:

> On Fri, 6 Aug 2004 06:34, Si Jones <si at bananas.hopto.org> wrote:

>> The first attempt of the first install went wrong as I reboot after
>> installing, and the raid array had not synced the mirror, the
>> second attempt I left it to sync and rebooted and it was fine.

> Anyway this doesn't sound like an anaconda bug.  Once the RAID device is 
> active then it's up to the kernel to manage it.  If on a reboot a RAID device 
> is not recognised then it sounds to me like a kernel bug not an anaconda bug.

Right.  There's a bug in bugzilla about the kernel marking RAID
devices that are scheduled to begin the initial synchronization, or
even that begin it, but that don't have time to complete it, as if
they had completed, such that they don't get fully synced if there's a
reboot in between.  Since mdadm creates RAID devices in a way that is
equivalent to creating a degraded array and then completing it with
redundant devices (as opposed to mkraid's approach that created
complete raid devices just in need of a resync), this bug hits when
using mdadm to create the raid device.  This is a very serious bug
IMHO, but it has never been looked into, to the best of my knowledge.

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=116685

-- 
Alexandre Oliva             http://www.ic.unicamp.br/~oliva/
Red Hat Compiler Engineer   aoliva@{redhat.com, gcc.gnu.org}
Free Software Evangelist  oliva@{lsd.ic.unicamp.br, gnu.org}