At 08:21 PM 3/30/00 +0000, you wrote:
>Date: Thu, 30 Mar 2000 12:01:03 -0700
>From: Eric Sisler <[EMAIL PROTECTED]>
>To: [EMAIL PROTECTED]
>Subject: Re: Better backup plan? Suggestions or comments - anyone?
>Message-Id: <[EMAIL PROTECTED]>
>Content-Type: text/plain; charset="us-ascii"; format=flowed
>
>Steven Hildreth <[EMAIL PROTECTED]> wrote:
>
>>Hi, so here is the deal I am wanting to backup my two servers onto a single
>>tape backup.
>[...]
>1) NFS mout (read-only) / on the server to be backed up.
>[...]
>2) Backup the server without the tape drive using /etc/rmt and either rsh
>or ssh.
>[...]
>There's obviously some setup required for each of these methods, but I've
>experimented with both of them so yell if you want to try one.
>
>>[...]
>Yes, Linux has a 2Gb file size limitation at this time.
>
>>I want this backup to be automated. To be able to just replace tapes daily
>>and have crontabs ( as I do now) do the work and send me email reports of
>>the outcome.
>>[...]
>>Is this possible without the purchase of a third-party backup solution?
>
>Yes (with a few caveats). I happen to use BRU, which is a commercial
>backup product, but it should be doable with tar, cpio or other backup
>software and/or a different tape drive. It's really the methodology that's
>important, not the tools or hardware involved.
I too use BRU (the free version, from when Redhat included it in the 5.x
series, you're allowed as long as you've went thru the upgrade path
continuously), but I only backup my main "box" (that does everything
from samba/nfs to www & sendmail for the time being).
This past weekend I picked up some 486dx266's that are making excellent
linux boxes for all sorts of things, so I am trying to figure out how
to consistently back them all up too, with the convenience of my
nightly bru perl scripts running like they normally do.
It was my understanding that my version of BRU wouldn't backup
nfs mounted drives, so this weekend I'll find out whether I need
to purchase it or not, or move back to tar.
>Now for the caveats:
>In theory (and probably in reality) it's doable with a floppy disk, but I
>decided to go a slightly different route.
>[...]
In the distant past I've read posts by people saying that they've made good
boot floppies for use w/ bru on machines that are or are not direct/remote from
the
tape backup unit. I think I caught mentions of it on the http://www.estinc.com
discussion/listserv archives at http://www.estinc.com/cgi-bin/messages?list=bru
My thought was that the disk would have to have:
1) a kernel that can handle the geometry of all the hard drives
on all of the pc's you're backing up (ie if you're using
one of those 40gb maxtors you need the ide patch or a ~2.4ish
kernel.
2) fdisk, mke2fs, e2fsck, mt, as well as /bru, /etc/bru*
3) network config for eth0 as well as nic drivers
I found info about Tom's root/boot disk at http://www.toms.net/rb/
but haven't had time yet to read about it, nor the bootdisk howto
if one exists.
>I haven't had time to write up a good working description of how to go
>about using my method, but I know it works because I've used it. In short,
>it goes like this:
>
>1) If necessary, re-partition the disk drive.
>2) Install a *very* skeletal working copy of Linux to a "recovery"
>partition
>[...]
>3) Install/configure any other services needed for backup (eg. backup
>software, networking (if required for the restore), etc.)
>4) Recreate the necessary directory structure of the original server,
>mounting other filesystems to somewhere temporary, like /mnt/restore or the
>like.
>5) Restore from tape using /mnt/restore as the / directory.
>6) Run lilo to update the MBR. (There are specific switch settings for this.)
>7) Reboot and if all goes well, the server will be back to its backed-up
state.
>8) wipe the data from the temporary recovery partition (optional).
>
>Obviously there are specific things for each step that need to be followed,
>but that's the process in a nutshell. I've been threatening to write up
>the whole mess as a web page, but I haven't gotten around to it yet.
Oh, please do! Even though I think my method is heading into a
different direction then yours I'm betting I could benefit from
your documentation.
My general hope was to have run a script each night that would
1) mount all the machines on the tape backup server via nfs as ro
2) backup the following from each machine (if needed)
/home
/root
/bin
/bru
/etc
/lib
/sbin
/usr
/var
I'm pretty sure it can skip /proc because that'd be recreated when the
restored machine boots. I'm unsure about /dev.
Then on each machine, have a known machinename and ip for the "restore"
bootdisk. If it's booted, it can read from the tape and restore
a pc's drive(s).
3) then unmount the nfs'd partitions
Then with the boot floppy, on any machine I should be able to do a
partial or full restore, or complete rebuild if needed.
Any comments appreciated.
Shane
--
To unsubscribe: mail [EMAIL PROTECTED] with "unsubscribe"
as the Subject.
