CentOS - Jun 2010 - md0 and moving to another system

Hi all,

I thought to have tested this a while ago but I guess not.

I have a Centos 5.3 system (call its SYSA) software raid 0, 2 disks.

I have an identical system (SYSB) else were, so I pull one of my  
mirrors drives from SYSA and put it in SYSB as the only drive in SYSB.

Upon but I get grub error.

I boot in recuse mode (CD, linux rescue) and it attempts to mount the  
disk, but claims no partitions exists.

I drop into a shell and mount it w/o issues.

Running fdisk on the drive tells me its got 2 partitions which is true.

I set this up as ext3, no LVMs, etc... with root and the other  
partition being swap.

I've always thought that all I can do in the event of motherboard  
failure os to have a duplicate system around and merely move a dive  
off to it.

- aurf

>I thought to have tested this a while ago but I guess not.
Ok...
>I have a Centos 5.3 system (call its SYSA) software raid 0, 2 disks.
So a strip? Did you mean Raid 1 (Mirror)?
>I have an identical system (SYSB) else were, so I pull one of my  
>mirrors drives from SYSA and put it in SYSB as the only drive in SYSB.
>
>Upon but I get grub error.
So I am sure you meant Raid 1, otherwise I suspect you see the futility
in this?

You need to install grub to both discs, whereas the installer just hits
the first disc. In my MD Raid kickstart for systems like this I %post a
grub script that works as you describe:

# Install grub to both discs
/sbin/grub --batch >/root/grub_install.log <<EOF
root (hd0,0)
setup (hd0)
root (hd1,0)
setup (hd1)
quit
EOF

I actually use this on a dev system, I pull one of the two discs, boot and
do my tests then rebuild from the outdated original, works well...

jlc