Re: Anaconda, Grub, RAID 1, /boot

[Curtis Doty <Curtis GreenKey net>]

1:59pm Jack Neely said:

> On Fri, Feb 16, 2007 at 11:36:06AM -0500, seth vidal wrote:
> > On Fri, 2007-02-16 at 11:31 -0500, Jack Neely wrote:
> > > Folks,
> > > 
> > > I've run into something odd, and a way to repair the damage, but was
> > > hoping someone could shed some light on what's actually happening.
> > > 
> > > We've stabilized on RHEL 4.4 for the time being.  I have had several
> > > reports of folks installing a RAID 1 system with /boot mirrored and
> > > complained that Grub hangs or otherwise wont boot the system.  I've done
> > > this a lot and it always works for me.
> > > 
> > > Until I installed a new Dell 1950.  Grub tries to load stage 1.5 and
> > > runs off into random data land.
> > > 
> > > I used a Grub CD and 
> > > 
> > > root (hd0,0)
> > > setup (hd0)
> > > 
> > > if I was lucky, I would get different random data.  However, when I used
> > > my grub CD and the configfile option to bring up the system proper I can
> > > run a grub shell from the system and do this:
> > > 
> > > device (hd0) /dev/sda
> > > root (hd0,0)
> > > setup (hd0)
> > > 
> > > I reboot the system and it works like a champ.
> > > 
> > > Try to install Grub again using the first method with the grub CD and we
> > > are back to random data land.
> > > 
> > > What's happening here?  Is this #217176?
> > 
> > I've got a number of 1950s installed with /boot on sw-raid1. No problems
> > so far. Have you been able to verify this on multiple 1950s? Could it
> > just be this one?
> > 
> > -sv
> > 
> 
> I'm very lucky to have gotten this one...so I don't have other 1950s to
> test with.  Mind if I come pick up a few more?
> 
> What really makes me scratch my head is why my grub cd doesn't fix the
> MBR properly the first time around.  It is a newer version of grub than
> what's in RHEL 4.  *shrug*
> 

Seth, have you ever tried booting from the second disk?

I've seen this intermittently since forever... But have not yet come up 
with a clear answer/solution:

	http://bugzilla.redhat.com/168660

As you can see from my comments, there's an ugly hack if you are 
installing with kickstart. And to be safe, I also always run something 
like this post-install:

# dd if=/dev/sda count=1 |strings |grep GRUB
1+0 records in
1+0 records out
512 bytes (512 B) copied, 4.3e-05 seconds, 11.9 MB/s
GRUB 
# dd if=/dev/sdd count=1 |strings |grep GRUB 
1+0 records in
1+0 records out
GRUB 
512 bytes (512 B) copied, 3.7e-05 seconds, 13.8 MB/s

It's not perfect. But it does catch the times when the second disk never 
caugh the proper MBR. As no GRUB would appear.

When this happens, I *have* seen anaconda/grub fail with the "error 16" on 
the debugging tty.

If both disks were zeros before the install, then I can rest assured the 
second disk will still be bootable in a few years when the primary fails 
while I'm up in the mountains skiing or whatever... :-/

../C