UNIX for Dummies Questions & Answers

Hi,

I am creating a disaster recovery plan for my Linux 7.2 machine. I have two backups from my current machine.

One created using the command
tar -cvpf /dev/st0 --exclude=/proc --directory / .

and one created with the command
find / /boot /home -mount -path '/proc' -prune -o -print | cpio -ovcB -O /dev/st0

When I restore the "find" version everything works.

When I restore the "tar" version, the machine doesn't boot into Linux at all. It displays "LI" at the bottom, in white letters, and locks up.

The "find" method takes about 7 hours to run and the "tar" method takes 2.5 hours.

Can anyone tell me what is different about these two commands and what they back up, and why the big time difference?


thanks

I'm not sure why one takes so much longer than the other, but I've seen that problem on bootup (hangs at LI).
In that case, you boot manager/area is damaged - in this case, LILO.

To restore, I would recommend building a barebones system, then unpacking your backups onto it. Or you could take a disk image... It may be as easy as "dd if=/dev/hda of=/dev/st0" from a boot disk... Note that it may also be more complicated than that, and the device names would have to be adjusted (I'm assuming you have your entire normal filesystem on one EIDE disk... this is probably not the case).

If you're trying to make the existing backup work, you can unload it to disk, boot off of a disk (CDROM, Floppy, etc) with the root being your newly-restored disk, and execute "/sbin/lilo" to try to repair it...

Hope this helps a little...

thanks livinfree,

I am doing the bare bones system, (on a duplicate machine) and unpacking the backups.

Is there some reason that the boot manager area is not backed up or damaged with the "tar" method as opposed to the "find" method? Both backup methods are done at the init 1 level.

I'm very interested in the original posters question of why find|cpio restores without corruption LILO as does tar (seemingly).

From your post, it would seem that one corrupts (or indirectly influences the corruption or lack of something) of either the Master Boot Record (MBR) or the LILO boot maps.

If we assume that both find|cpio and tar archive and restore the LILO configurations files in the LILO configuration directory, then a logical (but wrong, me thinks) assumption is un-tar corrupts the MBR, and that does not seem possible.

The premise of the question is very interesting and has me contemplating how we could uncover the problem.??.. OR, if there is someother factor that is causing the LILO error unrelated to either find|cpio or tar.

I have found the reason for the big time difference between the two commands. If I increase the block-size parameter on the "find" method it takes about 45 minutes to do the backup.

new command:
find / /boot /home -mount -path '/proc' -prune -o -print | cpio -ovc -O /dev/st0 --block-size=40

per NEO, I am in the process of comparing the LILO configuration files to determine if something is causing problems there.

Thanks for all the advice.