How To Image SUSE Linux Systems Using ``tbku`` ============================================== This document describes how to use the TundraWare Inc. ``tbku`` utility to "image" or "clone" SUSE Linux systems. .. Note:: Most/Much of this will also be relevant to other Linux distributions, though some of the fine points may be different. Why Bother Imaging? =================== Suppose we need to build a new instance of a ``SUSE Linux`` system. Perhaps we need to replace one that just had a hard drive failure. Maybe we want to build a new server that is based on our "standard" system configuration. In other words, we want to go from "bare metal" hardware to a fully running *and configured* system as quickly as possible. There are a number of commercial and open source solutions to this problem, but they all have one thing in common: We want to minimize the amount of manual labor needed to install, configure, and otherwise customize the final system. This is especially important in large data centers where it is impractical to manually (re)install each and every server, its applications, and its customization information. "Imaging" or "Cloning" allows us to keep a copy of the entire OS *as configured* - that means with all its applications and configuration options set up as desired. We then load or "Provision" a new hard drive with this image and *voila'*, "instant" running system. What Is ``tbku``? ================= ``tbku`` is a shell script that makes it easy to create tarballs of some of all of your filesystems. ``tbku`` does not help you with *restoring* your image, it's just handy for creating the image in the first place. If you've never used it before, take a moment to download it and read the documentation. You'll find the latest copy at: http://www.tundraware.com/Software/tbku There is no fee for using ``tbku`` in any context, personal or commercial. However, there are some licensing terms you have to abide by to use it, so take a moment to read the license in the distribution tarball. .. Note:: You don't *have* to use ``tbku`` to create your backup image. The description below should work fine so long as you have a backup of all the relevant files that preserves all the appropriate file information such as ownership and permissions. ``tbku`` just makes it easy to automate the creation of such backups. The Big Picture =============== Before diving into the details, it's good to get a sense of the overall process. Imaging a system requires the following steps: A. Create the master image: - Create a baseline system configured as you want it. - Take an "image" of it. (That's where ``tbku`` is helpful.) - Save the image somewhere (DVD, USB drive, network drive ...) you can get at when you need it to (re)install a system. B. Use the master image to (re)provision a machine: - Prepare the target hard disk to receive the image. - Dump the image onto the hard disk. - Adjust the configuration if/as needed for the new hardware. Creating The Master Image ========================= Unlike other approaches that make an image of *the disk*, ``tbku`` creates an image of *files* on the disk. This means that your new target disk does not have to be physically the same as the one on which the master image (sometimes called a "snapshot") was made. You can clone systems back and forth between SCSI, IDE, and SATA. You can clone from smaller disks to larger ones or go the other way. .. Note:: The whole point of doing imaging is to avoid having to do custom configuration for each new installation. However, some configuration changes may be necessary when the target hardware is different than the hardware on which the master image was created. This is discussed a bit more below in the `Gotchas`_ section. Creating The Master Image 1. Select the machine whose existing SUSE Linux installation you want preserved or used as a standard installation image. 2. Image that system with ``tbku`` using the following fileset:: /bin /boot /dwnl /etc /home /lib /local /opt /root /sbin /srv /usr /var Notice that we do *not* backup the dynamic kernel-created filesystems like ``/dev`` or ``/proc``. 3. Save the resulting ``.tar.gz`` (tarball) file somewhere it can be retrieved later when you want to image another machine. This can be a network server, a USB drive, a DVD or whatever makes sense in your environment. As with all backup systems, it's pretty important to make multiple copies of the backup image, and keep a couple of them off-site. Provisioning With The Master Image ================================== Now that we have a "snapshot" or master image, we can use it to (re)provision machines. Provisioning Machines With A Master Image 1. Boot the SUSE Linux installation disk and load the ``Rescue System``. 2. Now we have to prepare the disk to receive a Linux filesystem. The example below assumes we are installing on /dev/hda - a PATA master on the first IDE controller - but you can do what follows with any of the drives on your system. Just substitute the device names as appropriate:: # Partition the drive: fdisk /dev/hda # Delete and create partitions as you like # Make sure the partition that will mount / # is toggled to be bootable # Be sure to use the 'write' option before exiting # Suppose you end up with this: # # /dev/hda1 is for swap (type 82) # /dev/hda2 is for your filesystem (type 83) mkswap /dev/hda1 mkfs.reiserfs /dev/hda2 mount /dev/hda2 /mnt # Now, let's create the top level directories that that # were not backed up and/or will be used by the kernel # for its own filesystems: cd /mnt mkdir dev media mnt proc sys tmp # Now it's time to mount your backup medium. Depending # on your backup medium this can be one of several # devices. CD/DVDs are often found at /dev/hdc1. USB # drives show up as SCSI drives such as /dev/sda1, and # so on. You'll also need to know the type of the # backup medium (see: man mount for the details): mount -tvfat /dev/sda1 /mnt/mnt # This is a USB drive # OK, time to dump the image previously created by tbku # onto our shiny new filesystem (make sure your current # directory is still /mnt before doing this): tar -xzvf /mnt/my-system-image.tar.gz # Now we have to make sure that the boot tables and # default file mounts are correct - Our target system # may have a different drive type or device (SCSI, sata, # pata) than the system from which tbku took the image: # Edit /mnt/etc/fstab to reflect the partitioning # you did with fdisk. Remember that drives can be # named by device name (/dev/xxxx) or by the vendor # name (/dev/disk/by-id/xxxx). In our case the # relevant portion of /mnt/etc/fstab looks like this: /dev/hda1 swap swap defaults 0 0 /dev/hda2 / reiserfs acl,user_xattr 1 1 # Be sure not to disturb the other stuff in the fstab # file # Now, check and fix the device map file, # /mnt/boot/grub/device.map Say we took the tbku image # from a system that boot from SCSI, that file # would look like this: (fd0) /dev/fd0 (hd0) /dev/sda # But our new system wants to boot from ATAPI so it # now needs to look like this: (fd0) /dev/fd0 (hd0) /dev/hda # We also have to correct any differences in the boot # menu that appears when you first start the system. # This is in /mnt/boot/grub/menu.lst Near the top of # this file you'll see something like this: gfxmenu (hd0,1)/boot/message # hd0 is right - we made sure of that when we edited # the map file above. Make sure that the offset (1 in # this case) is also right. This is the number, # *counting from 0* of the root/boot partition within that # drive. In our case, (hd0,1) is correct because our # root/boot partition is /dev/hda2. # Following this are the individual menu entries. # Make sure you check each line of every entry. # Suppose we find this: title SLED 10 - 2.6.16.54-0.2.5 root (hd0,1) kernel /boot/vmlinuz-2.6.16.54-0.2.5-default root=/dev/sda2 resume=/dev/sda1 splash=silent showopts initrd /boot/initrd-2.6.16.54-0.2.5-default # All references to /dev/sda2 have to be changed to /dev/hda2 # All references to /dev/sda1 have to be changed to /dev/hda1 # Repeat this for every menu entry. # Finally, lets make sure that the boot loader # is properly installed and configured: grub-install --root-directory=/mnt /dev/hda We're DONE! Well ... maybe. If the hardware (chipset, CPU) of your target machine is similar/same as the machine from which you took the original image OR if the kernel you plan to boot has support for your new target hardware, you should just be able to boot and run at this point. If not, read the following `Gotchas`_ section for further explanation. This may all seem complex the first time you do it, but after a couple of times, you'll be able to do this in your sleep. Depending on how large your backup image is, a complete system restore can typically be done in less than an hour. That's less than an hour to a *completely configured system* with all your applications, custom configuration, and so on as you last left them. Gotchas ======= There are some circumstances where you cannot avoid doing some configuration on the newly provisioned machine. A. Network differences B. Different Hardware C. Different Chip Architectures Author ====== Tim Daneliuk - tbku@tundraware.com Comments and/or improvements welcome! Document Information ==================== This document produced using the amazing ``reStructuredText`` tools in the ``docutils`` package. For more information, see: http://docutils.sourceforge.net/rst.html ``$Id: Imaging-SUSE-Linux-With-tbku.txt,v 1.104 2008/03/12 22:40:44 tundra Exp $``