ADSM-L

Re: Need Linux help to create disaster recovery instructions....

2002-08-07 15:18:37
Subject: Re: Need Linux help to create disaster recovery instructions....
From: Thomas Denier <Thomas.Denier AT MAIL.TJU DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Wed, 7 Aug 2002 15:22:55 -0400
> I am trying to write bare metal recovery instructions for Linux.
>
> Has anyone else done this?
>
> I am trying a couple different methods....
>
> 1) Install a minimal redhat system on one partition, then create needed
> partitions of system being restored, and then restoring "in place" (will red
> hat choke if system files are overwritten while redhat is running?)..
>
> 2) Install minimal redhat system, create partitions, and mount on temp
> mounts points (/tmp/root, /tmp/usr/, /tmp/var, etc)  so as not to interfere
> with system. Update fstab/mtab/MBR and then reboot.....
>
> Anyone have any insight, experience with such things...?

I ran some experiments using an approach slightly different from either
of the ones mentioned above. I used an NFS server containing a copy of
part of the directory tree on a Red Hat installation CD and a copy of
TSM client files from a fully installed Linux system. I followed
instructions accompanying the CD to create a bootable diskette suitable
for starting an installation from an NFS server. I booted from the
diskette, told the kernel where the NFS server was, and requested a shell
prompt. I partitioned the hard disk, formatted the partitions, and set
up a hierarchy of mount points like /mnt/root, /mnt/root/usr, and so on.
I then restored / to /mnt/root, /usr to /mnt/root/usr, and so on. I
ran a command to rebuild the boot record on the hard disk and rebooted
from the hard disk.

I later tried putting a mutant version of Red Hat's install script on
the NFS server. The script prompted me for information it needed and
ran the necessary commands. I almost had this debugged when we started
using a new release of Red Hat with a new install script. I have since
concluded that my co-workers and I could not realistically plan on keeping
pace with Red Hat if we tried to use mutant install scripts as the basis
for our disaster recovery procedures.

I would have liked to simply boot from the CD, obtaining only the TSM
client files from an NFS server. I never figured out how to get networking
started with a kernel booted from the installation CD.

Our hot site vendor does not supply us with replacement hardware
identical to the hardware we normally use. I never got as far as sorting
out all the things that would need to be adjusted to get a Linux system
restored from TSM backups to run on one of the hot site machines.