Does this theory hold water

Mark Musone mmusone at shatterit.com
Wed Oct 16 10:51:18 EDT 2002


The other thing that is usually faster/easier is do not even bother with
NFS and just do the
untar/dump via ssh/rsh..no need to mount it if you just restore through
a network pipe..

(granted, it helps a lot more with actually doing the backups via
rsh/ssh than restoring, but restoring works
Just as well..)

-Mark


-----Original Message-----
From: owner-nflug at nflug.org [mailto:owner-nflug at nflug.org] On Behalf Of
Darin Perusich
Sent: Wednesday, October 16, 2002 10:36 AM
To: nflug at nflug.org
Subject: Re: Does this theory hold water


this should work, i've done similare restores on solaris systems. you 
shouldn't have to copy the tarball onto the freshly installed systems. 
before saying OK to reboot the system jump into another terminal 
(ctrl+alt+F3), ifconfig eth0 up, startup portmap and nfs. mount your 
network drive, and extract the tarball, lilo the disk and reboot.

this is something i'd test and document all the steps before a fire 
starts, knowing that you've done it before can save alot of agrivation 
and get you home earlier.

Justin Bennett wrote:
> Ok let me know if this makes sense I know there has been some 
> discussion on here in the past, I've never done this but think it 
> should work.
> 
> I have a RedHat 6.2 Pro web server, I'm running a nightly tar:
> 
> tar -czf /home/backup/backup.tar.gz * --directory / --exclude=proc 
> --exclude=home/backup --exclude=mnt --exclude=*/lost+found
> 
> to a network drive, I like that because the tar is only 900M and if 
> our Webmaster blows something away (you know those graphic artist 
> types) :) it's easy to pull something out of a tar on disk rather than

> tape. if this drive were to fail. I should be able to recover this 
> way:
> 
> 1. Reload box with redhat 6.2
>    Use same partition scheme
> 2. Boot into new 6.2
> 3. Copy tar to local drive
> 4. Boot into rescue mode (so that no files are used/locked on local
disk)
>    mknod /dev/sda1,2,3, ect
>    Mount disk structure under something like /mnt/root
>    chroot /mnt/root
> 5. Dump tar to disk (overlaying files)
> 6. Run lilo
> 7. Reboot and boot from disk
> 
> Does this make sense? Do I need to do all this or can I boot into 
> rescue mode, fdisk and mkfs on all drives, mount and chroot, dump tar,

> lilo, and reboot. I was just wondering if I don't do an install if 
> things like /mnt are made? Sounds like it might be fun to try on a 
> test box... :)
> 
> Thanks
> Justin
> 


-- 
Darin Perusich
Unix Systems Administrator
Cognigen Corp.
darinper at cognigencorp.com






More information about the nflug mailing list