ccollect - Restoring backups ============================ Nico Schottelius 0.1, for all ccollect version, Initial Version from 2008-07-04 :Author Initials: NS Having backups is half the way to success on a failure. Knowing how to restore the systems is the other half. Introduction ------------ You made your backup and now you want to restore your data. If you backuped only parts of a computer and need only to restore them, it is pretty easy to achieve. Restoring a whole system is a little bit more difficult and needs some knowledge of the operating system. Restoring parts of a system ~~~~~~~~~~~~~~~~~~~~~~~~~~~ Log into your backupserver. Change into the backup directory you want to restore from. Do `rsync -av './files/to/be/recovered/' 'sourcehost:/files/to/be/recovered/'. Restoring a complete system (general) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Boot the system to be rescued from a media that contains low level tools for your OS (like partitioning, formatting) and the necessary tools (ssh, tar or rsync). Use - create the necessary partition table (or however it is called Get a live-cd, that ships with - rsync / tar - ssh (d) -> from backupserver - support for the filesystems Restoring a complete FreeBSD system ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Get a FreeBSD-live-cd (I used the FreeBSD 7.0 live CD, but FreeSBIE (http://www.freesbie.org/), Frenzy (http://frenzy.org.ua/en/) or the FreeBSD LiveCD (http://livecd.sourceforge.net/) may also be helpful. The following way uses the FreeBSD 7.0 live cd. So boot it up, select your language. After that select *Custom* then *Partition*. Create the slice like you want to have it. Then let the installer write into the MBR, select *BootMgr*. After that create the necessary labels, select *Label* and make sure "Newfs" flag is set to "Y". Finally, select *Commit* and choose an installation type that must fail, because we want the installer only to write the partitions and labels, but not to install anything on it. At this point we have created the base for restoring the whole system. Move back to the main menu and select *Fixit*, then *CDROM/DVD*. This starts a shell on TTY4, which can be reached by pressing *ALT+F4*. Then enter the following data: -------------------------------------------------------------------------------- rootdir=/ccollect rootdev=/dev/ad0s1a backupserver=192.42.23.5 # create destination directory mkdir "$rootdir" # mount root; add other mounts if you created more labels mount "$rootdev" "$rootdir" # find out which network devices exist ifconfig # create the directory, because dhclient needs it mkdir /var/db # retrieve an ip address dhclient fxp0 # test connection ssh "$backupserver" # go back backupserver% exit -------------------------------------------------------------------------------- Now we've prepared everything for the real backup. The next problem maybe, that we cannot (should not) be able to login as root to the backup server. Additionally the system to be restored may not reachable from the backup server, because it is behind a firewall or nat. Thus I describe a way, that is a little bit more complicated for those, that do not have these limitations, but works in both scenarios. I just start netcat on the local machine, pipe its output to tar and put both into the background. Then I create a ssh tunnel to the backupserver, which is then able to connect to my netcat "directly". -------------------------------------------------------------------------------- # user to connect to the backupserver myuser=nico # our name in the backup restorehost=server1 # the instance to be used backup="weekly.20080718-2327.23053" # Need to setup lo0 first, the livecd did not do it for me ifconfig lo0 127.0.0.1 up # change to the destination directory cd "$rootdir" # start listener ( nc -l 127.0.0.1 4242 | tar xvf - ) & # verify that it runs correctly sockstat -4l # connect as a normal user to the backupserver ssh -R4242:127.0.0.1:4242 "$myuser@$backupserver" # become root backupserver% su - # change to the source directory backupserver# cd /home/server/backup/$restorehost/$backup # begin the backup backup # tar cf - . | nc 127.0.0.1 4242 # wait until it finishes, press ctrl-c to kill netcat # logoff the backupserver backupserver# exit backupserver% exit -------------------------------------------------------------------------------- Now we are just right next to be finished. Still, we have to take care about some things: - Do the block devices still have the same names? If not, correct /etc/fstab. - Do the network devices still have the same names? If not, correct /etc/rc.conf. If everything is fixed, let us finish the restore: -------------------------------------------------------------------------------- # cleanly umount it umount "$rootdir" # reboot, remove the cd and bootup the restored system reboot -------------------------------------------------------------------------------- Restoring a complete Linux system ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Knoppix knoppix 2 at boot prompt rootdir=/ccollect dev=/dev/hda rootdev="${dev}1" fs=jfs tar # create the needed partitions cfdisk $dev mkfs.$fs $rootdev mkdir $rootdir mount $rootdev $rootdir cd $rootdir pump ifconfig # start listener (from now on it is the same as ( nc -l 127.0.0.1 4242 | tar xvf - ) & TO BE DONE Future ------ I think about automating full system recoveries in the future. I think it could be easily done and here are some hints for people who would like to implement it.