tbku - Table-driven backup script
``tbku allsets | [fileset] ...``
``tbku`` is a utility script for producing "tarball" backups of
some- or all of your files. It is useful both for producing
incremental backups or for systemwide images or "snapshots". The
script can be run either from the command line or, more typically,
as a ``cron`` job to automate system backup tasks.
``tbku`` uses standard utilities common on Unix-like systems, like
``tar``, ``sed``, and ``uname``. It uses no other special or custom
tools. For this reason, it is highly portable across many variants
of these systems.
The central benefit of using ``tbku`` over hand written ``tar``
commands is that ``tbku`` is "table driven". You specify the set
of files to back up in a table (a separate file). You can have as
many of these "filesets" as you wish, corresponding to different
kinds of backups you want done. ``tbku`` will do backups
automatically or manually, based on the name of the "fileset". This
considerably simplifies automating backups, keeping backup logs, and
generally maintaining an orderly backup environment.
``tbku`` was originally developed as a backup tool for FreeBSD
servers. Since then, it has been updated to also work with SUSE
Linux, both servers and desktops. ``tbku`` should work with little-
or no modification on any other Unix-like system. For example,
``tbku`` will run without modification (other than default
locations) in a ``cywgin`` environment under MS-Windows.
To use ``tbku``, all you have to do is install the file somewhere in
your ``$PATH``. Typically, a good place for it is in
``/usr/local/bin``. Just make sure its permissions are 755 so all
users will be able to use it.
You may optionally want to put ``tbku.1.gz`` somewhere in your
``$MANPATH`` so this documentation will be available as a man page.
There is also a ``tbku`` port for FreeBSD users that automates the
installation and deinstallation of ``tbku``. This port can be found
under ``/usr/ports/sysutils/tbku``. Once installed, all of the
documentation for ``tbku`` (in a variety of formats) including the
tool itself, the licensing terms, and the instructions for imaging
systems with it, will be found in ``/usr/local/share/doc/tbku``.
Once you've installed the program, you should verify that its
default settings are to your liking. If not, you can override them
via environment variables (described later in this document). For
interactive use, make sure the environment variables you want to set
are exported when you log in. If you're running ``tbku`` from a
``cron`` job, be sure to set the environment variables of interest
in the ``crontab`` file.
``tbku`` has to know just *what* you want backed up. You do
this by creating a so-called *fileset* in the appropriate
directory (default: ``$HOME/tbku/``). Filesets are just text
files that list all the files and/or directories that are
to be backed up together. For instance, suppose you had
a fileset called ``manual.fileset.homedirs`` that contained
just these three lines::
If you now run this command::
The files and/or contents of ``/root``, ``/home``, and ``/usr/home``
would be written to a tarball in the backup directory (default:
``/bku/``). By default, the resulting tarball's name has a long
string of text that includes the machine name, system type, OS type,
date, *and* the so-called *set name*. The "set name" is nothing
more than the suffix of the name of the fileset used to produce the
tarball, in this case, ``homedirs``.
Additionally, you'll also find a log of the backup and "dot files"
that tell you when the backup began and when it ended. Here's part
of what you might see if you did an ``ls -a /bku``::
The "dot files" don't actually contain any information, but their
date/time stamps (you can see this with ``ls -al /bku``) will tell
you when the backup began and ended.
The log file contains a list of all the files that actually
made it into the tarball. The log file also captures *the
errors* encountered during a backup. This means that ``tbku``
is generally pretty quiet during a backup run. It scribbles
any complaints it has into the log. So... you should check
your logs regularly to make sure everything is working as
You can create as many different filesets as you like (for as many
different kinds of backups as you need). So, for example, you may
have one for the files you want backed up daily, another for weekly
backups, another for taking a snapshot of the entire system, and so
The *name* of a fileset can be used to change ``tbku`` behavior
(described below). The *content* of a fileset file must conform
to only a few rules:
1) Each line may contain the name of a *single* file or directory.
You cannot place multiples of these on a single line.
2) Each entry should be *an absolute path*. That way, ``tar``
will be able to figure out what it is you want to back up. By
default, most modern ``tar`` implementations will strip the
leading ``/`` so your backup tarball will be relative to
wherever you are when you restore from it.
3) There is no support for comments or other metadata inside a
fileset. File- and directory names are the *only* thing
that should ever be there.
``tbku`` semantics (behavior) depend on how you've named your
filesets. In general, a fileset should be named as follows::
Any fileset name that begins with "auto." will automatically be
backed up when you run the script without arguments::
If a fileset begins with something other than "auto.", you
have to explicitly name the set on the command line for
it to be backed up. Say we have only two filesets, ``manual.fileset.music``
and ``manual.fileset.docs``. Then::
tbku # Does nothing
tbku music # Only backs up the manual.fileset.music fileset
tbku music docs # Backs up both filesets
The "setname" is used to uniquely name each backup tarball.
Strictly speaking, ``tbku`` only cares about the "auto" string.
Anything other than "auto" as a prefix in the fileset name, will
cause the file to be seen as requiring manual invocation. Using
"manual" is just a helpful convention.
Similarly, you don't need the "fileset" in the middle of the
filename, it's just a helpful convention. ``tbku`` only examines
the prefix of the filename (up to the ".") to determine whether to
do automatic backups. It uses the suffix (from the last "." to the
end of the file name) to determine the set name. In fact, you
don't even have to fully specify the set name, just any trailing
tbku ic # Backs up manual.fileset.music
While these little semantic subtleties may be interesting, you are
strongly *discouraged* from using them, as they are not guaranteed
to be preserved in future releases of ``tbku``. Stick to the
conventions described above, and you should be fine.
**The allsets Option**
As you might guess, you can force *all* backup sets to be done
regardless of whether they are marked as "auto" or "manual"
by doing this::
The "allsets" argument must be the first argment on the command
line, and anything following it will be ignored. In other words,
only the form shown above is meaningful.
**tbku: Nothing to do!**
You may see ``tbku`` grumbling about having nothing to do. This
happens under one of several circumstances:
1) You ran ``tbku`` without arguments, but there are no
"auto" filesets defined.
2) You ran ``tbku`` with arguments, but no filesets with
matching set names were found.
3) There are no filesets at all.
**Autodeletion Of Old Backups**
As shipped, ``tbku`` uniquely identifies each backup set based on
machine name, OS, CPU architecture, set name, and, most importantly,
date. If you've set it up to run as a cron job, over time you'll
accumulate lots of older copies of backups. That's because each new
day, the backup file name will change (since it includes the date).
If you don't like this default behavior, change the ``TBKUDEL``
environment variable to be "YES". It must be *exactly* this string,
all in upper case. Anything else will cause ``tbku`` to *not*
autodelete old backups. This is intentional, to make it hard to
accidentally enable this feature.
Enabling this feature forces ``tbku`` to delete all older files
associated with the selected set name. This includes the start/stop
"dot" files, the log, and the backup tarball itself. In effect,
this option forces ``tbku`` to only keep the most recent backup of
each backup set.
*Use this option with caution!* If you only keep the most recent
copy of your backups in your backup directory, you may never be able
to get to changes made days, weeks, or months prior.
**IMAGING WITH tbku**
It is possible to use ``tbku`` backups to completely (re)image a
machine. The general idea is to have ``tbku`` produce a tarball of
all the (relevant) files on the system you want to "clone". Then,
you can dump that onto a newly prepared filesystem on the target
machine. This is a handy (and relatively quick) way to recover a
system after a hard drive failure or upgrade, for example.
The ``tbku`` distribution contains separate documents that describe
in detail how to image both FreeBSD and SUSE Linux systems. You can
also read the documents on line at:
``tbku`` is written to be "smart" enough to figure out where your
system keeps needed tools like ``tar`` or ``sed``. The only
requirement here is that ``tbku`` be run in an environment that can
find the ``which`` command somewhere in its ``$PATH`` - ``tbku``
uses ``which`` to figure out just where everything it needs lives on
your filesystem. If ``tbku`` cannot figure out where your system
keeps things, it will use the FreeBSD default values.
For most FreeBSD and Linux users, this should work without any
customization beyond setting environment variables to override
default behavior (described below). In rare circumstances, you may
need further customization. All the things you're likely to ever
want to change appear first in the actual ``tbku`` script, and are
briefly documented there.
**DEFAULTS & ENVIRONMENT VARIABLES**
You can override the various ``tbku`` defaults by setting a
corresponding environment variable.
================= =============================== =========================
**Env. Variable** **Default Value** **Meaning**
----------------- ------------------------------- -------------------------
TBKUDEL NO YES -> Delete old backups
TBKUDIR /bku Where to write backups
TBKUNAME $MACHINE-$OSTYPE-$OSREV-$HWTYPE Tarball base name
TBKUSETS $HOME/tbku Filesets found here
TBKUTAPE /dev/sa0 Tape device (or file)
================= =============================== =========================
export TBKUDIR=/mnt/backups # Backups written to /mnt/backups
export TBKUNAME=JoeBackup # Backups named: JoeBackup-<setname>
export TBKUSETS=/tbku # Looks for filesets in /tbku
export TBKUTAPE-/tmp/faketape # Tape backups actually written to *file*
export TBKUDEL="YES" # Autodelete old backups when starting a set
``tbku`` was originally designed for use by experienced
systems administrators and users. As such, it does little
or no error checking. If you define backup or fileset
directories that are non-existent, for instance, you will
get strange behavior. ``tbku`` *will* try to create the
backup directory you've specified if it does not already
exist, but this may not work if you're running as anything
other than ``root`` user.
``tbku`` is intended to make it easier/more automatic to
to backups. It is not, however, idiot-proof. There are
some general backup guidelines you should observe:
**NEVER, EVER, EVER, EVER, EVER ... EVER**, trust a backup tool
until you've confirmed that it is correctly producing backups
**and** you can properly restore from them!
Always keep multiple copies of your backups. If ``tbku`` is
writing its backups to the same drive/system it runs on, **make
sure you also keep a copy of those backups "off system"**.
It's a pretty good idea to keep **multiple backup copies**, on
**different media** (disk, tape, DVD, thumbdrive), in **different
**UPDATES & SUPPORT**
To get the latest version of 'tbku', go to:
For questions, comments, or other feedback, send email to:
Tim Daneliuk, TundraWare Inc.
**COPYRIGHT & LICENSING**
``tbku`` is Copyright (c) 2004-2008, TundraWare Inc., Des Plaines, IL, USA
There is no fee for using ``tbku`` either personally or commercially
*so long as the terms of the tbku license are met*. Please read the
``tbku-license.txt`` file for a full explanation of the licensing
This document was produced using the very useful
``reStructuredText`` tools in the ``docutils`` package. For more
``$Id: tbku.txt,v 1.110 2008/03/20 17:26:46 tundra Exp $``