Guide pratique de sauvegarde et de récupération sous Linux

Ce processus de sauvegarde vous oblige à disposer du lecteur de disque ZIP à chaque sauvegarde. Vous pouvez créer le contenu du disque ZIP dans un répertoire, puis le sauvegarder via le réseau. Ensuite, il vous faudra simplement construire un disque ZIP sur le serveur de sauvegarde (avec la commande cp -rp) lorsque vous aurez besoin de restaurer.

Le processus de sauvegarde sera plus rapide, mais vérifiez tout de même que le répertoire créé tiendra sur votre disque ZIP (avec la commande du -hs $cible.zip) ! Vous devrez éditer la définition de la variable zip dans le fichier save.metadata.

Mon portable ne supportant pas la présence simultanée d'une carte réseau et d'un lecteur ZIP, c'est le processus que j'utilise pour le sauvegarder. Je conserve une image de sauvegarde en même temps que la courante, ce qui fait que je dispose d'un système de secours au cas où l'ordinateur planterait pendant une sauvegarde.

Vous pouvez aussi sauvegarder sur plusieurs disques ZIP des données qui en valent la peine, et les transférer sur le système pendant la restauration.

Ceci est similaire à l'option « pas de lecteur ZIP » ci-dessus. Enregistrez vos sauvegardes dans un répertoire de votre disque dur, comme mentionné. Puis, utilisez la commande mkisofs pour créer une image ISO 9660 de ce répertoire et gravez-la.

De nos jours, beaucoup d'ordinateurs sont livrés avec un lecteur de CDROM mais pas de disquette. De plus, les lecteurs de disquette tombent en panne. C'est pourquoi, graver votre CDROM avec une image qui permette d'amorcer le système est une bonne idée. La mauvaise nouvelle, c'est que le format « El Torito » supporte les disquettes de 1,2 Mo, 1,44 Mo et 2,88 Mo, et que tomsrtbt utilise une disquette de 1,7 Mo. La bonne nouvelle, c'est que vous pouvez obtenir une version en 2,88 Mo, tomsrtbt-2.0.103.ElTorito.288.img , des mêmes miroirs que là où vous vous procurez l'image pour disquette. Placez une copie ^[1] dans le répertoire racine des fichiers de sauvegarde. Puis, utilisez l'option -b de la commande mkisofs pour spécifier que tomsrtbt-2.0.103.ElTorito.288.img est le fichier de l'image d'amorçage.

Le seul point noir de ce processus est que beaucoup de vieux BIOS ne supportent pas les images de disquette de 2,88 Mo sur CDROM. Beaucoup de ceux-ci s'amorceront avec une disquette tomsrtbt.

On peut aussi utiliser Syslinux. Il n'est pas dépendant de l'image d'une disquette et vous pouvez construire votre propre CD avec un certain nombre d'outils tels que tomsrtbt.

Il vous faudra peut-être ajuster les options du BIOS pour pouvoir amorcer sur le lecteur de CDROM.

Testez vos CDROM sur le lecteur que vous utiliserez pour la restauration. S'il se trouve que vous devez modifier les scripts, vous pouvez les copier dans /tmp, un disque en mémoire vive sous tomsrtbt, et les y éditer. Les scripts s'exécuteront à cet emplacement. Un disque en MEV étant volatil par nature, enregistrez vos changements avant de réamorcer !

En séparant les deux premiers scripts de la première étape, restore.metadata et save.metadata, vous pouvez répartir les méta données de la première étape sur plusieurs disques ZIP.

Lors de la première étape, il est parfois nécessaire de compresser quelques méga-octets de données, particulièrement lorsque vous atteignez les limites de votre disque ZIP. La fonction crunch du script save.metadata accepte de nombreux paramètres pour alimenter la commande tar. Elle accepte aussi le paramètre --exclude. Ainsi, par exemple, vous pouvez exclure les répertoires samba et X11 situés dans /etc de cette façon :

crunch etc etc --exclude etc/samba --exclude etc/X11

Pourquoi ces deux-là ? Parce qu'ils consomment beaucoup d'espace disque et que nous n'en avons pas besoin pour l'amorçage.

Si vous possédez un nombre important de noyaux, vous pouvez éliminer les modules de tous les noyaux sur lesquels vous n'amorcerez pas. vérifiez votre lilo.conf ou grub.conf pour connaître le noyau que vous utiliserez, puis vérifiez /lib/modules pour connaître les répertoires des modules que vous pouvez exclure.

Où trouver d'autres bons candidats à l'exclusion ? Listez les répertoires cible avec la commande ls -alSr pour les fichiers individuels, et avec la commande du | sort -n pour les répertoires.

Une autre façon (probablement plus nette), d'exclure les répertoires est d'écrire la liste complète des répertoires dans un fichier puis d'y faire référence via l'option tar --exclude-from=FILENAME.

Si votre système utilise un disque MEV pour initialiser, ou initrd pour amorcer, assurez-vous que restore.metadata crée le répertoire /initrd. La façon la plus simple de le faire est de s'assurer qu'ils figurent dans la liste des répertoires utilisés dans la boucle de création des répertoires (vers la fin).

Votre système utilise probablement initrd s'il démarre sur un disque SCSI ou si sa racine est sur une partition ext3fs. Vérifiez dans /etc/lilo.conf s'il appelle un tel script.

Le processus automatique lance tous les scripts manuels dans le bon ordre. Il ne permet pas les interventions manuelles, telles que la création de systèmes de fichiers que ce guide pratique ne prend pas en charge. Pour lancer la première étape de restauration automatiquement, saisissez :

# /mnt/root.bin/first.stage

Si vous désirez rechercher les blocs endommagés, ajoutez l'option « -c ».

Pour lancer le processus manuel, positionnez-vous dans le répertoire où se trouvent les scripts, dans le lecteur ZIP.

# cd /mnt/root.bin

Lancez maintenant le(s) script(s) qui restaurer(a/ont) les informations de partition et créer(a/ont) les systèmes de fichiers. Vous pouvez les lancer dans n'importe quel ordre, par exemple :

# ./make.dev.hda

Si vous désirez qu'une recherche de blocs endommagés s'effectue, ajoutez l'option « -c ».

Ce script va :

	N.B.
	Si vous devez restaurer d'autres systèmes d'exploitation, c'est le moment de le faire. Lorsque c'est terminé, redémarrez avec tomsrtbt et continuez la restauration de Linux.

À présent, lancez le(s) script(s) qui crée(nt) des points de montage et montez les partitions.

# ./mount.dev.hda

Une fois que vous avez créé tous vos répertoires et que vous y avez monté les partitions, vous pouvez lancer le script restore.metadata . Il restaurera le contenu du lecteur zip sur le disque dur.

Vous devriez voir un répertoire du répertoire racine du lecteur zip, puis une liste des fichiers d'archives tels qu'ils sont restaurés. L'exécution de la commande tar sous tomsrtbt vous dira que la taille d'un bloc de tar est de 20, et c'est très bien ainsi. Vous pouvez l'ignorer. Assurez-vous que lilo affiche ses résultats :

Added linux *

Un affichage suivra, correspondant à la commande « df -m ».

Si d'habitude vous démarrez directement sous X, vous aurez peut-être quelques problèmes. Pour les éviter, changez temporairement votre niveau de démarrage. Dans le fichier /target/etc/inittab, trouvez la ligne qui ressemble à :

id:5:initdefault:

et transformez-la en :

id:3:initdefault:

Vous pouvez maintenant démarrer en douceur. Retirez la disquette tomsrtbt de votre lecteur si vous ne l'avez pas encore fait, et faites le salut à trois doigts à votre ordinateur (N.D.T. : CTRL+MAJ+SUP), ou son équivalent :

# reboot

L'ordinateur va s'arrêter et redémarrer.

Ce script, lancé pendant la sauvegarde, crée des scripts semblables à make.dev.hda et mount.dev.x, plus bas, pour que vous les lanciez à la restauration. Il produit aussi des fichiers de données semblables à dev.hda, plus bas. Le nom du script et du fichier de données qui est produit dépend du périphérique donné en paramètre à ce script. Ce script, lancé à la restauration, crée les partitions sur le disque dur. make.fdisk est appelé par save.metadata, plus bas.

#! /usr/bin/perl

# A perl script to create a script and input file for fdisk to
# re-create the partitions on the hard disk, and format the Linux and
# Linux swap partitions. The first parameter is the fully qualified
# path of the device of the hard disk, e.g. /dev/hda. The two
# resulting files are the script make.dev.x and the data file dev.x
# (where x is the hard drive described, e.g. hda, sdc). make.dev.x is
# run at restore time to rebuild hard drive x, prior to running
# restore.metadata. dev.x is the input file for fdisk.

# Time-stamp: <2004-04-10 13:51:37 root make.fdisk>

# Copyright 2001 through the last date of modification Charles Curley
# except for the subroutine cut2fmt.

# cut2fmt Copyright (c) 1998 Tom Christiansen, Nathan Torkington and
# O'Reilly & Associates, Inc.  Permission is granted to use this code
# freely EXCEPT for book publication.  You may use this code for book
# publication only with the explicit permission of O'Reilly &
# Associates, Inc.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# In addition, as a special exception, Tom Christiansen, Nathan
# Torkington and O'Reilly & Associates, Inc.  give permission to use
# the code of this program with the subroutine cut2fmt (or with
# modified versions of the subroutine cut2fmt that use the same
# license as the subroutine cut2fmt), and distribute linked
# combinations including the two.  You must obey the GNU General
# Public License in all respects for all of the code used other than
# the subroutine cut2fmt.  If you modify this file, you may extend
# this exception to your version of the file, but you are not
# obligated to do so.  If you do not wish to do so, delete this
# exception statement and the subroutine cut2fmt from your version.

# You can also contact the Free Software Foundation at
# http://www.fsf.org/

# Changes:

# 2004 04 10: fdisk v > 2.11 has wider columns. Added code to select
# the appropriate cut string based on fdisk's version.

# 2004 04 09: Added support for Mandrake's idea of devfs. On Mandrake,
# everything is mounted with devfs. So the mount devices are buried
# deep in places like /dev/ide/host0/bus0/target0/lun0/part1 instead
# of places like /dev/hda1, where $DEITY intended they should be. We
# have to reverse from the long devfs device to the shorter old style
# that tomsrtbt uses. The alternative is to keep track in an array of
# which devfs device belongs to which short device.

# 2003 12 29: Changed the regex for detecting whether a file system is
# read-write in the code that builds the mount file(s). The old test
# does not work if mount returns multiple parameters in the 5th field,
# e.g. (rw,errors=remount-ro) on some debian systems. This regex
# assumes that the rw parameter is always listed first, which may not
# always be the case. If it fails, take out the '\('. Thanks to Pasi
# Oja-Nisula <pon at iki dot fi> for pointing this out.

# 2003 01 09: Added support for FAT32. We now create two scripts for
# each hard drive, make.dev.[as]dx and mount.dev.[as]dx. These create
# and make file systems on each partition, and make mount points and
# mount them.

# 2002 12 25: added support to handle W95 extended (LBA) (f) and W95
# FAT 32 partitions. I have tested this for primary but not logical
# partitions.

# 2002 09 08: Added minimal support for ext3fs. We now detect mounted
# ext3fs partitions & rebuild but with no options. The detection
# depends on the command line "dumpe2fs <device> 2>/dev/null | grep -i
# journal" producing no output for an ext2fs, and output (we don't
# care what) for an ext3fs.

# This could stand extension to support non-default ext3 options such
# as the type of journaling. Volunteers?

# 2002 07 25: Bad block checking is now a command line option (-c) at
# the time the product script is run.

# 2002 07 03: Corrected the mechanism for specifying the default
# drive.

# 2001 11 25: Changed the way mke2fs gets its bad block
# list. badblocks does not guess at the block size, so you have to get
# it (from dumpe2fs) and feed it to badblocks. It is simpler to just
# have mke2fs call badblocks, but you do loose the ability to have a
# writing test easily. -- C^2

# 2001 11 25: Changed the regex that extracts partition labels from
# the mount command. This change does not affect the results at all,
# it just makes it possible to use Emacs' perl mode to indent
# correctly. I just escaped the left bracket in the regex. -- C^2

# Discussion:

# fdisk will spit out a file of the form below if you run it as "fdisk
# -l".

# root@tester ~/bin $ fdisk -l /dev/hda

# Disk /dev/hda: 64 heads, 63 sectors, 1023 cylinders
# Units = cylinders of 4032 * 512 bytes

#    Device Boot    Start       End    Blocks   Id  System
# /dev/hda1             1         9     18112+  83  Linux
# /dev/hda2            10      1023   2044224    5  Extended
# /dev/hda5            10       368    723712+  83  Linux
# /dev/hda6           369       727    723712+  83  Linux
# /dev/hda7           728       858    264064+  83  Linux
# /dev/hda8           859       989    264064+  83  Linux
# /dev/hda9           990      1022     66496+  82  Linux swap

# What fdisk does not do is provide output suitable for later
# importing into fdisk, a la sfdisk. This script parses the output
# from fdisk and creates an input file for fdisk. Use the input file
# like so:

# fdisk /dev/hdx < dev.hdx

# For the bare metal restore package, this script also builds a script
# that will execute the above command so you can run it from your zip
# disk. Because the bare metal restore scripts all are in /root/bin,
# the data file and script created by this script are also placed
# there. The same script also creates appropriate Linux file systems,
# either ext2fs, or Linux swap. There is limited support for FAT12,
# FAT16 and FAT32. For anything else, you're on your own.

# Note for FAT32: According to the MS KB, there are more than one
# reserved sectors for FAT32, usually 32, but it can vary. Do a search
# in M$'s KB for "boot sector" or BPB for the gory details. For more
# info than you really need on how boot sectors are used, see
# http://support.microsoft.com/support/kb/articles/Q140/4/18.asp

# You can also edit dev.x to change the sizes of partitions. Don't
# forget, if you change the size of a FAT partition across the 32MB
# boundary, you need to change the type as well! Run "fdisk /dev/hda"
# or some such, then the l command to see the available partition
# types. Then go ahead and edit dev.x appropriately. Also, when moving
# partition boundarys with hand edits, make sure you move both logical
# and extended partition boundaries appropriately.

# Bad block checking right now is a quick read of the partition. A
# writing check is also possible but more difficult. You have to run
# badblocks as a separate command, and pass the bad block list to
# mke2fs in a file (in /tmp, which is a ram disk). You also have to
# know how large the blocks are, which you learn by running
# dumpe2fs. It gets messy and I haven't done it yet. You probably
# don't need it for a new hard drive, but if you have had a hard drive
# crash on you and you are reusing it (while you are waiting for its
# replacement to come in, I presume), then I highly recommend it. Let
# me know how you do it.

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.


# cut2fmt figures out the format string for the unpack function we use
# to slice and dice the output from fdisk. From Christiansen and
# Torkington, Perl Cookbook 5.

sub cut2fmt {
    my (@positions) = @_;
    my $template    = '';
    my $lastpos     = 1;

    foreach $place (@positions) {
        $template .= "A" . ($place - $lastpos) . " ";
        $lastpos = $place;
    }

    $template .= "A*";
    return $template;
}


# Sub gpl, a subroutine to ship the GPL and other header information
# to the current output file.

sub gpl {

print OUTPUT <<FINIS;

# Copyright 2001 through the last date of modification Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.

FINIS

}


# Begin main line code.



# Provide a default device.

# print "\$ARGV[0] is $ARGV[0].\n";

$device = defined ($ARGV[0]) ? $ARGV[0] : "/dev/hda";

# Need to check to see if $device is a sym link. If it is, the mount
# point is the target of the link. (Mandrake) Otherwise we search for
# mount points on $device. Fedora, Red Hat.

if ( -l $device) {

    # It is a sym link. Get the target of the link, then make it into
    # an absolute path, preserving the numbering.

    $mountdev = '/dev/' . readlink ($device);
    $mountdev =~ s|ide/host(\d+)/bus(\d+)/target(\d+)/lun(\d+)/disc|ide/host\1/bus\2/target\3/lun\4|;
} else {
    # not a sym link; just assign it.
    $mountdev = $device;
}

# print "Device is $device; mount device is $mountdev.\n";

# Prepare format string. Here are two format strings I have found
# useful. Here, column numbers are 1 based, i.e. the leftmost column
# is column 1, not column 0 as in Emacs.

# We select a format string according to fdisk's version.

$fdpid = open (FDVER, "fdisk -v |") or die "Couldn't fork: $!\n";
while (<FDVER>) {
    @_ = unpack ("A7 A*", $_);
    $fdver=$_[1];
    $fdver =~ s/[^\d.]//g; # strip non-numbers, non-periods, as in "2.12pre".
}

# print "fdisk version is $fdver\n";

if ($fdver < 2.12) {
# fdisk to 2.11?? Red Hat, Fedora Core 1
    $fmt = cut2fmt (11, 19, 24, 34, 45, 49);
} else {
# fdisk 2.12 & up?? Mandrake 10.0, Fedora Core 2
    $fmt = cut2fmt (12, 14, 26, 38, 50, 55);
}
# print "Format string is $fmt.\n";

# define fields in the array @_.
$dev = 0;
$bootable = 1;
$firstcyl = 2;
$lastcyl = 3;
$parttype = 5;
$partstring = 6;

$target = "\/target";

$outputfilename = $device;
$outputfilename =~ s/\//./g;
$outputfilename = substr ($outputfilename, 1, 100);

$outputfilepath = "\/root\/bin\/";


# Make a hash of the labels.
$mpid = open (MOUNT, "mount -l |") or die "Couldn't fork: $!\n";
while (<MOUNT>) {
    if ($_ =~ /^$mountdev/i) { # is this a line with a partition in it?
#       print $_;               # print it just for grins
        split;
        if ($_[6] ne "") {      # only process if there actually is a label
            $_[6] =~ s/[\[\]]//g; # strike [ and ].
            $labels{$_[0]} = $_[6];
#           print "The label of file device $_[0] is $labels{$_[0]}.\n";
        }


        # We only mount if it's ext2fs or ext3fs and read and write.

        if ($_[4] =~ /ext[23]/ and $_[5] =~ /\(rw/ ) {
            if ($_[0] =~ /ide/i) {

                # We have a devfs system, e.g. Mandrake. This code
                # converts back from the devfs designation to the old
                # /dev/hd* designation for tomsrtb. I have NOT checked
                # this out for drives other than /dev/hda. Also, this
                # code does not handle SCSI drives.

                if ( $_[0] =~ /target0/ && $_[0] =~ /bus0/ ) {
                    $letter = 'a';
                } elsif ( $_[0] =~ /target1/ && $_[0] =~ /bus0/) {
                    $letter = 'b';
                } elsif ( $_[0] =~ /target0/ && $_[0] =~ /bus1/) {
                    $letter = 'c';
                } else {
                    $letter = 'd';
                }
                $_[0] =~ s|/ide/host\d+/bus\d+/target\d+/lun\d+/part|/hd|g;
                $_[0] =~ s/hd/hd$letter/;
            }
            $mountpoints{$_[2]} = $_[0];
#             print "$_[2] is the mountpoint for tomsrtbt device $mountpoints{$_[2]}.\n";
        }
    }
}
close (MOUNT);


$fpid = open (FDISK, "fdisk -l $device |") or die "Couldn't fork: $!\n";

open (OUTPUT, "> $outputfilepath${outputfilename}")
    or die "Couldn't open output file $outputfilepath${outputfilename}.\n";

while (<FDISK>) {
    if ($_ =~ /^$device/i) {    # is this a line with a partition in it?
#       print $_;               # print it just for grins
        chop;                   # kill trailing \r
        @_ = unpack ($fmt, $_);

        # now strip white spaces from cylinder numbers
        @_[$firstcyl] =~ s/[ \t]+//;
        @_[$lastcyl] =~ s/[ \t]+//;
        @_[$parttype] =~ s/[ \t]+//;

        $partnumber = substr(@_[$dev], 8, 10); # get partition number for this line
        # just for grins
#       print "  $partnumber, @_[$firstcyl], @_[$lastcyl], @_[$parttype], @_[$partstring]\n";

        # Here we start creating the input to recreate the partition
        # this line represents.

        print OUTPUT "n\n";
        if ($partnumber < 5) {
            # primary Linux partition
            if (@_[$parttype] == 83) {
                print OUTPUT "p\n$partnumber\n@_[$firstcyl]\n";
                if (@_[$firstcyl] ne @_[$lastcyl]) { # in case it's all on one cylinder
                    print OUTPUT "@_[$lastcyl]\n";
                }

                # Now detect if this is an ext3 (journaling)
                # partition. We do this using dumpe2fs to dump the
                # partition and grepping on "journal". If the
                # partition is ext2, there will be no output. If it is
                # ext3, there will be output, and we use that fact to
                # set a command line switch. The command line switch
                # goes into an associative array (hash) so we don't
                # have to remember to reset it to the null string when
                # we're done.

                $dpid = open (DUMPE2FS, "dumpe2fs @_[$dev] 2>/dev/null | grep -i journal |")
                    or die "Couldn't fork: $!\n";
                while (<DUMPE2FS>) {
#                   print "Dumpe2fs: $_";
                    $ext3{$_[$dev]} = "-j ";
                    last;
                }
                close (DUMPE2FS);

                if ($labels{@_[$dev]}) { # do we have a label?
                    $format .= "echo\necho formatting $checking@_[$dev]\n";
                    $format .= "mke2fs $ext3{$_[$dev]}\$blockcheck -L $labels{@_[$dev]} @_[$dev]\n\n";
                } else {
                    $format .= "echo\necho formatting $checking@_[$dev]\n";
                    $format .= "mke2fs $ext3{$_[$dev]}\$blockcheck @_[$dev]\n\n";
                }

                # extended partition
            } elsif (@_[$parttype] == 5) {
                # print ("Creating Extended Partition.\n");
                print OUTPUT "e\n$partnumber\n@_[$firstcyl]\n";
                if (@_[$firstcyl] ne @_[$lastcyl]) {
                    print OUTPUT "@_[$lastcyl]\n";
                }

                # extended partition, Win95 Ext'd (LBA)
            } elsif (@_[$parttype] eq "f") {
                # print ("Creating Extended LBA Partition.\n");
                print OUTPUT "e\n$partnumber\n@_[$firstcyl]\n";
                if (@_[$firstcyl] ne @_[$lastcyl]) {
                    print OUTPUT "@_[$lastcyl]\n";
                }
                print OUTPUT "t\n$partnumber\nf\n";

                # primary Linux swap partition
            } elsif (@_[$parttype] == 82) {
                print OUTPUT "p\n$partnumber\n@_[$firstcyl]\n";
                if (@_[$firstcyl] ne @_[$lastcyl]) {
                    print OUTPUT "@_[$lastcyl]\n";
                }
                print OUTPUT "t\n$partnumber\n82\n";
                $format .= "echo Making @_[$dev] a swap partition.\n";
                $format .= "mkswap \$blockcheck @_[$dev]\n\n";

                # Primary mess-dos partition. We don't handle hidden
                # partitions.
            } elsif ( @_[$parttype] == 1 || @_[$parttype] == 4 || @_[$parttype] == 6
                      || @_[$parttype] eq "b" || @_[$parttype] eq "c"
                      || @_[$parttype] eq "e" ) {
                # print ("Making DOS primary partition.\n");
                print ("dd if=@_[$dev] of=$outputfilepath$outputfilename$partnumber");
                print (" bs=512 count=1\n");
                system ("dd if=@_[$dev] of=$outputfilepath$outputfilename$partnumber bs=512 count=1");
                print OUTPUT "p\n$partnumber\n@_[$firstcyl]\n";
                if (@_[$firstcyl] ne @_[$lastcyl]) { # in case it's all on one cylinder
                    print OUTPUT "@_[$lastcyl]\n";
                }
                print OUTPUT "t\n$partnumber\n@_[$parttype]\n";
                $format .= "echo\necho formatting $checking@_[$dev]\n";
                $format .= "mkdosfs \$blockcheck";
                if ( @_[$parttype] == b || @_[$parttype] == c) {
                    # We have a W9x FAT32 partition. Add a command line switch.
                    $format .= " -F 32";
                }
                $format .= " @_[$dev]\n";
                $format .= "# restore FAT boot sector.\n";
                $format .= "dd if=$outputfilename$partnumber of=@_[$dev] bs=512 count=1\n\n";

            } else {
                # anything else partition
                print OUTPUT "p\n@_[$firstcyl]\n";
                if (@_[$firstcyl] ne @_[$lastcyl]) {
                    print OUTPUT "@_[$lastcyl]\n";
                }
                print OUTPUT "t\n$partnumber\n@_[$parttype]\n";
            }

        } else {
            # logical Linux partition
            if (@_[$parttype] == 83) {
                print OUTPUT "l\n@_[$firstcyl]\n";
                if (@_[$firstcyl] ne @_[$lastcyl]) {
                    print OUTPUT "@_[$lastcyl]\n";
                }

                # Now detect if this is an ext3 (journaling)
                # partition. We do this using dumpe2fs to dump the
                # partition and grepping on "journal". If the
                # partition is ext2, there will be no output. If it is
                # ext3, there will be output, and we use that fact to
                # set a command line switch. The command line switch
                # goes into an associative array (hash) so we don't
                # have to remember to reset it to the null string when
                # we're done.

                $dpid = open (DUMPE2FS, "dumpe2fs @_[$dev] 2>/dev/null | grep -i journal |")
                    or die "Couldn't fork: $!\n";
                while (<DUMPE2FS>) {
#                   print "Dumpe2fs: $_";
                    $ext3{$_[$dev]} = "-j ";
                    last;
                }
                close (DUMPE2FS);

                if ($labels{@_[$dev]}) { # do we have a label?
                    $format .= "echo\necho formatting $checking@_[$dev]\n";
                    $format .= "mke2fs $ext3{@_[$dev]}\$blockcheck -L $labels{@_[$dev]} @_[$dev]\n\n";
                } else {
                    $format .= "echo\necho formatting $checking@_[$dev]\n";
                    $format .= "mke2fs $ext3{@_[$dev]}\$blockcheck @_[$dev]\n\n";
                }

                # logical Linux swap partition
            } elsif (@_[$parttype] == 82 ) {
                print OUTPUT "l\n@_[$firstcyl]\n";
                if (@_[$firstcyl] ne @_[$lastcyl]) {
                    print OUTPUT "@_[$lastcyl]\n";
                }
                print OUTPUT "t\n$partnumber\n82\n";
                $format .= "echo Making @_[$dev] a swap partition.\n";
                $format .= "mkswap \$blockcheck @_[$dev]\n\n";

                # Logical mess-dos partition. We don't handle hidden
                # partitions.

            } elsif ( @_[$parttype] == 1 || @_[$parttype] == 4 || @_[$parttype] == 6
                      || @_[$parttype] eq "b" || @_[$parttype] eq "c"
                      || @_[$parttype] eq "e" ) {
#               print ("Making DOS logical partition.\n");
                print ("dd if=@_[$dev] of=$outputfilepath$outputfilename$partnumber");
                print (" bs=512 count=1\n");
                system ("dd if=@_[$dev] of=$outputfilepath$outputfilename$partnumber bs=512 count=1");
                print OUTPUT "l\n$partnumber\n@_[$firstcyl]\n";
                if (@_[$firstcyl] ne @_[$lastcyl]) { # in case it's all on one cylinder
                    print OUTPUT "@_[$lastcyl]\n";
                }
                print OUTPUT "t\n$partnumber\n@_[$parttype]\n";
                $format .= "echo\necho formatting $checking@_[$dev]\n";
                $format .= "mkdosfs \$blockcheck";
                if ( @_[$parttype] == b || @_[$parttype] == c) {
                    # We have a W9x FAT32 partition. Add a command line switch.
                    $format .= " -F 32";
                }
                $format .= " @_[$dev]\n";
                $format .= "# restore FAT boot sector.\n";
                $format .= "dd if=$outputfilename$partnumber of=@_[$dev] bs=512 count=1\n\n";

            } else {
                # anything else partition
                print OUTPUT "l\n@_[$firstcyl]\n";
                if (@_[$firstcyl] ne @_[$lastcyl]) {
                    print OUTPUT "@_[$lastcyl]\n";
                }
                print OUTPUT "t\n$partnumber\n@_[$parttype]\n";
            }

        }

        # handle bootable partitions
        if (@_[$bootable] =~ /\*/) {
            print OUTPUT "a\n$partnumber\n";
        }
    }
}

print OUTPUT "v\nw\n";

close (OUTPUT);
close (FDISK);


open (OUTPUT, "> ${outputfilepath}make.$outputfilename")
    or die "Couldn't open output file ${outputfilepath}make.$outputfilename.\n";

print OUTPUT <<FINIS;
#! /bin/sh

# A script to restore the partition data of a hard drive and format
# the partitions. Created at bare metal backup time by the Perl script
# make.fdisk.
FINIS

&gpl;

print OUTPUT <<FINIS;

export blockcheck=\$1;

if [ "\$blockcheck" != "-c" ] && [ -n "\$blockcheck" ]
then
    echo "\${0}: automated restore with no human interaction."
    echo "\${0}: -c: block check during file system making."
    exit 1;
fi

FINIS

# Clean the old partition table out.
print OUTPUT "dd if=/dev/zero of=$device bs=512 count=2\n\nsync\n\n";

print OUTPUT "fdisk $device \< $outputfilename\n\nsync\n\n";
print OUTPUT $format;

print OUTPUT "fdisk -l \"$device\"\n";

close (OUTPUT);

# Now build the script that will build the mount points on the root
# and other partitions.

open (OUTPUT, "> ${outputfilepath}mount.$outputfilename")
    or die "Couldn't open output file ${outputfilepath}make.$outputfilename.\n";

print OUTPUT <<FINIS;
#! /bin/sh

# A script to create a minimal directory tree on the target hard drive
# and mount the partitions on it. Created at bare metal backup time by
# the Perl script make.fdisk.
FINIS

&gpl;

print OUTPUT <<FINIS;

# WARNING: If your Linux system mount partitions across hard drive
# boundaries, you will have multiple "mount.dev.* scripts. You must
# ensure that they run in the proper order. The root partition should
# be mounted first, then the rest in the order they cascade. If they
# cross mount, you'll have to handle that manually.

FINIS


# We have a hash of mount points and devices in %mountpoints. However,
# we have to process them such that directories are built on the
# appropriate target partition. E.g. where /usr/local is on its own
# partition, we have to mount /usr before we build /usr/local. We can
# ensure this by sorting them. Shorter mount point paths will be built
# first. We can't sort a hash directly, so we use an array.

# We build commands to create the appropriate mount points and then
# mount the partitions to the mount points. This is in preparation for
# untarring the contents of the ZIP disk, done in restore.metadata.

foreach $point ( sort keys %mountpoints) {
    print OUTPUT "\n# $point is the mountpoint for tomsrtbt device $mountpoints{$point}.\n";
    print OUTPUT "mkdir $target$point\n";
    print OUTPUT "mount $mountpoints{$point} $target$point\n";
}

print OUTPUT "\nmount | grep -i \"$device\"\n";

close (OUTPUT);

# These scripts are dangerous & should only be visible to root.

chmod 0700, "${outputfilepath}make.$outputfilename";
chmod 0700, "${outputfilepath}mount.$outputfilename";
chmod 0600, "${outputfilepath}$outputfilename";

Ce script est un exemple de la production de make.fdisk, plus haut. Il fait appel à des fichiers de données comme dev.hda, plus bas. Il crée des partitions et des systèmes de fichiers sur certains d'entre elles. C'est le premier script lancé à la restauration.

Si vous êtes suffisamment courageux pour éditer dev.hda (voir plus haut), pour, disons, ajouter une nouvelle partition, il se peut que vous deviez éditer ce script.

Si vous voulez que make.dev.hda vérifie les secteurs défectueux quand il installe un système de fichiers sur les partitions, saisissez en ligne de commande l'option "-c".

#! /bin/sh

# A script to restore the partition data of a hard drive and format
# the partitions. Created at bare metal backup time by the Perl script
# make.fdisk.

# Copyright 2001 through the last date of modification Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.


export blockcheck=$1;

if [ "$blockcheck" != "-c" ] && [ -n "$blockcheck" ]
then
    echo "${0}: automated restore with no human interaction."
    echo "${0}: -c: block check during file system making."
    exit 1;
fi

dd if=/dev/zero of=/dev/hda bs=512 count=2

sync

fdisk /dev/hda < dev.hda

sync

echo
echo formatting /dev/hda1
mkdosfs $blockcheck /dev/hda1
# restore FAT boot sector.
dd if=dev.hda1 of=/dev/hda1 bs=512 count=1

echo
echo formatting /dev/hda2
mke2fs -j $blockcheck -L /boot /dev/hda2

echo
echo formatting /dev/hda3
mke2fs -j $blockcheck -L / /dev/hda3

echo Making /dev/hda5 a swap partition.
mkswap $blockcheck /dev/hda5

fdisk -l "/dev/hda"

Ce script est un exemple de la production de make.fdisk, plus haut. Il crée des points de montage et monte les partitions, préparant le système de fichiers cible à la restauration des fichiers. C'est le second script lancé à la restauration.

Si vous êtes suffisamment courageux pour éditer dev.hda (voir plus haut), pour, disons, ajouter une nouvelle partition, il se peut que vous deviez éditer ce script.

#! /bin/sh

# A script to create a minimal directory tree on the target hard drive
# and mount the partitions on it. Created at bare metal backup time by
# the Perl script make.fdisk.

# Copyright 2001 through the last date of modification Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.


# WARNING: If your Linux system mount partitions across hard drive
# boundaries, you will have multiple "mount.dev.* scripts. You must
# ensure that they run in the proper order. The root partition should
# be mounted first, then the rest in the order they cascade. If they
# cross mount, you'll have to handle that manually.


# / is the mountpoint for /dev/hda3.
mkdir /target/
mount /dev/hda3 /target/

# /boot is the mountpoint for /dev/hda2.
mkdir /target/boot
mount /dev/hda2 /target/boot

mount | grep -i "/dev/hda"

Ce fichier de données est utilisé au moment de la restauration. Il est utilisé par fdisk et est alimenté par make.dev.hda. Il est produit pendant la sauvegarde par make.fdisk. Ceux qui connaissent bien fdisk reconnaîtront que chaque ligne est une commande ou une valeur de fdisk, tel qu'un numéro de cylindre. Donc, il est possible de changer les tailles des partitions et d'ajouter de nouvelles partitions en éditant ce fichier. C'est pourquoi la pénultième commande est v, qui vérifiera la table des partitions avant qu'elle soit écrite.

n
p
1
1
29
t
1
6
a
1
n
p
2
30
44
n
e
3
45
1023
n
l
45
944
n
l
945
1023
t
6
82
v
w

C'est le premier script lancé dans le processus de sauvegarde. Il appelle make.fdisk, plus haut. Si vous devez sauvegarder un disque dur SCSI ou plusieurs disques durs, faites de sorte que l'appel à make.fdisk soit effectué de façon correcte.

#! /bin/sh

# A script to save certain meta-data off to the boot partition. Useful for
# restoration.

# Time-stamp: <2004-04-29 15:36:52 root save.metadata>

# Copyright 2000 through the last date of modification, Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/


# 2003 01 08: We now age the output from rpm -VA to make back
# comparisons easier.

# The loop that creates directories now has the -p option for mkdir,
# which means you can create parents on the fly if they don't already
# exist.

# initrd is now in the list of directories to create automatically.

# We now exclude more stuff when building the tarballs.

# 2002 07 01: Went to bzip2 to compress the archives, for smaller
# results. This is important in a 100MB ZIP disk. Also some general
# code cleanup.

# 2002 07 01: The function crunch will tar and BZIP2 the
# archives. This is cleaner than the old code, and has better safety
# checking.


# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.


# Crunch: A function to compress the contents of a directory and put
# the archive onto the ZIP disk.

# The first parameter is the name of the archive file to be
# created. The backup location, $zip, will be prepended and the
# extension, "tar.bz2" will be appended.

# All following parameters will be taken as additional directories or
# files to be put into the archive.

function crunch {

if [ -z "$1" ] || [ -z "$2" ]	# Checks if parameter #1 or #2 is zero length.
then
   echo "-Parameter #1 or #2 is missing.-"  # Also if no parameter is passed.
   return 1
else
   local file=$1		# The archive file to create
   shift			# Discard the file name
   local dirs=$@		# The director[y|ies] to archive
   local tarcmd="tar -cjf"	# The tar command.

   $tarcmd  $zip/$file.tar.bz2 $dirs # do it!!

   error=$?			# Preserve the exit code

   if [ $error != 0 ]		# Did we fail?
   then				# Yes
      echo "Tar failed with error $error"
      echo $tarcmd $zip/$file.tar.bz2 $dirs
      exit $error		# return tar's exit code as ours
   fi

   return 0			# For error testing if needed.
fi
}

# Begin the main line code
export zip="/mnt/zip";		# Where we will put archives
#  export save="/mnt/save";

RPMVABACKS=/etc			# where we keep our backups
RPMVAROOT=rpmVa			# The root name of the pg backups
ANC=${RPMVABACKS}/${RPMVAROOT}.anc	# name for the oldest (ancient) backup
OLD=${RPMVABACKS}/${RPMVAROOT}.old	# name for the middling oldest backup
NEW=${RPMVABACKS}/${RPMVAROOT}.txt	# name for the newest backup

if [ -f ${ANC} ]; then
echo "Deleting ${ANC}"
rm ${ANC}
fi

if [ -f ${OLD} ]; then
echo "Aging ${OLD}"
mv ${OLD} ${ANC}
fi

if [ -f ${NEW} ]; then
echo "Aging ${NEW}"
mv ${NEW} ${OLD}
fi


# Now we save hard drive information. Run make.fdisk on each hard
# drive in the order in which it mounted from the root partition. That
# is, run it first on the hard drive with your root partition, then
# any hard drives that mount to the first hard drive, then any hard
# drives that mount to those. For example, if your root partition is
# on /dev/sdc, run "make.fdisk /dev/sdc" first.

# The reason for this is that make.fdisk produces a script to make
# mount points and then mount the appropriate partition to them during
# first stage restore. Mount points must be created on the partition
# where they will reside. The partitions must be mounted in this
# order. For example, if your /var and /var/ftp are both separate
# partitions, then you must mount /, create /var, then mount /var,
# then create /var/ftp. The order in which the script "first.stage"
# runs the mounting scripts is based on their time of creation.

# If necessary, put a line, "sleep 1" between calls to make.fdisk.

echo "Saving hard drive info"
make.fdisk /dev/hda

# back up RPM metadata

echo "Verifying RPMs."

rpm -Va | sort +2 -t ' ' > ${NEW}

echo "Finished verifying RPMs; now mounting the ZIP drive."

# Make sure we have the ZIP drive mounted.
umount $zip
modprobe ppa			# Driver for 100MB parallel port ZIP disk
mount $zip			# It should have ext2fs on partition 1.

# clean it all out
rm -r $zip/*
mkdir $zip/lost+found

echo "`hostname` bare metal ZIP disk, created `date`. `uname -a`" > $zip/README.txt

echo "Building the ZIP drive backups."

# These are in case we need to refer to them while rebuilding. The
# rebuilding process should be mostly automated, but you never
# know....

fdisk -l /dev/hda > $zip/fdisk.hda

ls -al /mnt > $zip/ls.mnt.txt
ls -al / > $zip/ls.root.txt

cd /

# Build our minimal archives on the ZIP disk. These appear to be
# required so we can restore later on.

crunch boot boot
crunch root root --exclude root/.cpan --exclude root/.mozilla
crunch etc etc --exclude etc/samba --exclude X11
crunch lib lib

crunch usr.sbin usr/sbin
crunch usr.bin usr/bin --exclude usr/bin/emacs --exclude usr/bin/emacs-21.2 --exclude usr/bin/emacsclient --exclude usr/bin/emacs-nox --exclude usr/bin/gs --exclude usr/bin/pine --exclude usr/bin/gimp-1.2 --exclude usr/bin/doxygen --exclude usr/bin/postgres --exclude usr/bin/gdb --exclude usr/bin/kmail --exclude usr/bin/splint --exclude usr/bin/odbctest --exclude usr/bin/php --exclude usr/bin/xchat --exclude usr/bin/gnucash --exclude usr/bin/pdfetex  --exclude usr/bin/pdftex --exclude usr/bin/smbcacls --exclude usr/bin/evolution-calendar --exclude usr/bin/xpdf --exclude usr/bin/xmms
crunch sbin sbin
crunch bin bin
crunch dev dev
crunch kerberos usr/kerberos/lib/

# Now optional saves.

# arkeia specific:
# crunch arkeia usr/knox

# save these so we can use ssh for restore. *crack* for RH 7.0 login
# authentication.
crunch usr.lib usr/lib/*crack* usr/lib/libz* usr/lib/libssl* usr/lib/libcrypto*

# save the scripts we used to create the ZIP disk and the ones we will
# use to restore it.
mkdir $zip/root.bin
cp -p /root/bin/* $zip/root.bin
rm $zip/root.bin/*~ $zip/root.bin/#*#

echo "Testing our results."
find $zip -iname "*.bz2" | xargs bunzip2 -t

# Not a normal part of the process: we duplicate the ZIP disk onto an
# NFS mount elsewhere.

#  echo "Backing the ZIP drive to the NFS mount."

#  umount $save
#  mount $save

#  rm -r $save/zip
#  mkdir $save/zip
#  cp -pr $zip $save

du -hs ${zip}*
df -m

Ce script restaure les métadonnées du disque ZIP à la première étape de la restauration.

#! /bin/sh

# A script to restore the meta-data from the ZIP disk. This runs under
# tomsrtbt only after partitions have been rebuilt, file systems made,
# and mounted. It also assumes the ZIP disk has already been
# mounted. Mounting the ZIP disk read only is probably a good idea.

# Time-stamp: <2003-08-23 10:09:16 ccurley restore.metadata>

# Copyright 2000 through the last date of modification Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# 2003 08 23: Oops: tar on tomsrtbt does not respect -p. Try setting
# umask to 0000 instead.

# 2003 02 13: Tar was not preserving permissions on restore. Fixed
# that.

# 2002 07 01: Went to bzip2 to compress the archives, for smaller
# results. This is important in a 100MB ZIP disk. Also some general
# code cleanup.

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.

umask 0000

zip="/mnt";			# Where we mount the zip drive.
target="/target";		# Where the hard drive to restore is mounted.

ls -lt $zip			# Warm fuzzies for the user.

cd $target

# Restore the archived metadata files.
for archive in $( ls $zip/*.bz2 ); do
echo $archive
ls -al $archive
bzip2 -dc $archive | tar -xf -
done

# Build the mount points for our second stage restoration and other
# things.

# If you boot via an initrd, make sure you build a directory here so
# the kernel can mount the initrd at boot.

for dir in mnt/zip mnt/cdrom mnt/floppy mnt/imports proc initrd; do
mkdir -p $target/$dir
done

chmod a-w $target/proc		# Restore /proc's read-only permissions

# Restore the scripts we used to create the ZIP disk and the ones we will
# use to restore it. These should be the latest & greatest in case we had
# to do any editing during 1st stage restore.
cp -p $zip/root.bin/* $target/root/bin

# Now install the boot sector.
chroot $target /sbin/lilo -C /etc/lilo.conf

df -m

Ce script effectue complètement la première étape de la restauration sans intervention humaine.

Si vous désirez vérifier les secteurs défectueux pendant la création des systèmes de fichiers des partitions, utilisez l'option de ligne de commande « -c ».

#! /bin/sh

# A master script to run the other, detailed scripts. Use this script
# only if you want no human intervention in the restore process. The
# only option is -c, which forces bad block checking during formatting
# of the partitions.

# Time-stamp: <2003-04-24 10:03:07 ccurley first.stage>

# Copyright 2002 through the last date of modification Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.

export blockcheck=$1;

if [ "$blockcheck" != "-c" ] && [ -n "$blockcheck" ]
then
    echo "${0}: automated restore with no human interaction."
    echo "${0}: -c: block check during file system making."
    exit 1;
fi

cd /mnt/root.bin		# just in case we aren't already where we should be.

for drive in $( ls make.dev.* ); do
    ./$drive $blockcheck;
done


# WARNING: If your Linux system mount partitions across hard drive
# boundaries, you will have multiple "mount.dev.* scripts. You must
# ensure that they run in the proper order, which the loop below may
# not do. The root partition should be mounted first, then the rest in
# the order they cascade. If they cross mount, you'll have to handle
# that manually.

# The "ls -tr" will list the scripts in the order they are created, so
# it might be a good idea to create them (in the script save.metadata)
# in the order in which you should run them.

for drive in $( ls -tr mount.dev.* ); do
    ./$drive;
done

./restore.metadata

# People who are really confident may comment this line in.
# reboot

Ce script effectue une sauvegarde vers un autre ordinateur via un montage NFS. Vous pouvez l'adapter pour effectuer vos sauvegardes vers une bande ou d'autres supports.

#! /bin/sh

# Back up the entire system to another computer's drive. To make this
# work, we need a convenient chunk of disk space on the remote computer we
# can nfs mount as /mnt/save.

# Time-stamp: <2003-04-24 09:56:05 ccurley back.up.all>

# Copyright 2000 through the last date of modification Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.

save="/mnt/save"

# Make sure it's there
umount $save
mount $save

cd /

rm $save/tester.tar.old.gz
mv $save/tester.tar.gz $save/tester.tar.old.gz

# save everything except /mnt, /proc, and nfs mounted directories.

time tar cf - / --exclude /mnt --exclude /proc --exclude $save\
    | gzip -c > $save/tester.tar.gz

Ce script fait exactement la même chose que back.up.all mais il utilise ssh à la place de nfs.

#! /bin/sh

# Back up the entire system to another computer's drive. To make this
# work, we need a convenient chunk of disk space on the remote
# computer. This version uses ssh to do its transfer, and compresses
# using bz2. This means this script has to know more about the other
# computer, which does not make for good modularization.

# Time-stamp: <2003-04-24 09:56:52 ccurley back.up.all.ssh>

# Copyright 2000 through the last date of modification Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.

save="/backs/tester"
backup_server="charlesc"

# rotate the old backups. Do it all in one line to minimze authentication overhead.
ssh $backup_server "rm $save/tester.tar.old.bz2; mv $save/tester.tar.bz2 \
    $save/tester.tar.old.bz2"

# save everything except /mnt, /proc, and squid directories.

time tar cf - / --exclude /mnt --exclude /proc --exclude /var/spool/squid\
    | ssh $backup_server "bzip2 -9 > $save/tester.tar.bz2"

Vous utiliserez ce script de restauration si votre sauvegarde a été effectuée avec back.up.all.

#! /bin/sh

# A script to restore all of the data from an nfs mount. This is our final
# stage restore.

# Time-stamp: <2003-04-24 09:58:51 ccurley restore.all>

# Copyright 2000 through the last date of modification Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.

export save="/mnt/save"

mount $save

cd /
gunzip -dc $save/tester.tar.gz | tar -xpkf -

rm /var/run/*.pid

lilo

Vous utiliserez ce script de restauration si votre sauvegarde a été effectuée avec back.up.all.ssh.

#! /bin/sh

# A script to restore all of the data using ssh and bunzip2. This is
# our final stage restore.

# Copyright 2000 through the last date of modification Charles Curley.

# Time-stamp: <2003-04-24 09:59:10 ccurley restore.all.ssh>

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.

save="/backs/tester/"
backup_server="charlesc"

cd /

ssh $backup_server "cat $save/tester.tar.bz2" | bunzip2 | tar -xpkf -

rm /var/run/*.pid

lilo

#! /bin/sh

# Back up another computer's drive to this system. To make this work, we
# need a convenient chunk of disk space on this computer. This version
# uses ssh to do its transfer, and compresses using bz2. This version was
# developed so that the system to be backed up won't be authenticated to
# log onto the backup computer. This script is intended to be used on a
# firewall. You don't want the firewall to be authenticated to the backup
# system in case the firewall is cracked.

# Time-stamp: <2004-04-03 12:24:12 ccurley get.tester>

# Copyright 2000 through the last date of modification Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.

# 2004 04 03: added /sys to the list of excludes. It is a read-only
# pseudo-file system like /proc.

# 2002 07 01: We now set the path on the target to the zip drive with
# a variable. This fixes a bug in the command to eject the zip disk.

# 2002 07 01: The zip disk archives are now in bzip2 format, so this
# script has been changed to reflect that.



# The host name of the computer to be backed up.
target=tester
zip=/mnt/zip

echo Backing up $target

echo Aging the ZIP disk backups.

rm -r $target.old.zip

mv $target.zip $target.old.zip

ssh $target "modprobe ppa ; mount -r $zip"

echo Copying the ZIP disk.

# -r for recursive copy, -p to preserve times and permissions, -q for
# quiet: no progress meter.

scp -qpr $target:$zip $target.zip

du -hs $target.zip


echo Aging the archives

rm $target.tar.old.bz2

mv $target.tar.bz2 $target.tar.old.bz2


echo Backing up $target to the backup server.

ssh $target tar -cf - / --exclude /sys --exclude /mnt --exclude /proc\
    --exclude /var/spool/squid\
    | bzip2 -9 | cat > $target.tar.bz2

echo Testing the results.
find . -iname "*.bz2" | xargs bunzip2 -t

ssh $target "eject $zip"

#! /bin/sh

# A script to restore all of the data to tester via ssh. This is our final
# stage restore.

# Time-stamp: <2003-04-24 09:59:45 ccurley restore.tester>

# Copyright 2000 through the last date of modification Charles Curley.

# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.

# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.

# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

# You can also contact the Free Software Foundation at http://www.fsf.org/

# For more information contact the author, Charles Curley, at
# http://www.charlescurley.com/.

# The host name of the computer to be restored.

target=tester

bunzip2 -dc $target.tar.bz2 | ssh $target "cd / ; tar -xpkf - "

ssh $target lilo