In general, you will need the Bacula source release, and if you want to run a Windows client, you will need the Bacula Windows binary release. However, Bacula needs certain third party packages (such as MySQL, PostgreSQL, or SQLite to build and run properly depending on the options you specify. Normally, MySQL and PostgreSQL are packages that can be installed on your distribution. However, if you do not have them, to simplify your task, we have combined a number of these packages into three depkgs releases (Dependency Packages). This can vastly simplify your life by providing you with all the necessary packages rather than requiring you to find them on the Web, load them, and install them.
The basic installation is rather simple.
If you are installing from rpms, and are using MySQL, please be sure to install mysql-devel, so that the MySQL header files are available while compiling Bacula. In addition, the MySQL client library mysqlclient requires the gzip compression library libz.a or libz.so. If you are using rpm packages, these libraries are in the libz-devel package. On Debian systems, you will need to load the zlib1g-dev package. If you are not using rpms or debs, you will need to find the appropriate package for your system.
Note, if you already have a running MySQL or PostgreSQL on your system, you can skip this phase provided that you have built the thread safe libraries. And you have already installed the additional rpms noted above.
SQLite is not supported on Solaris. This is because it frequently fails with bus errors. However SQLite3 may work.
make distclean
so that you are sure to start from scratch and not have a mixture of the two options. This is because ./configure caches much of the information. The make distclean is also critical if you move the source directory from one machine to another. If the make distclean fails, just ignore it and continue on.
If you skip this step (make) and proceed immediately to the make install you are making two serious errors:
make uninstall make distclean ./configure <your-new-options> make make install
If all goes well, the ./configure will correctly determine which operating system you are running and configure the source code appropriately. Currently, FreeBSD, Linux (Red Hat), and Solaris are supported. The Bacula client (File daemon) is reported to work with MacOS X 10.3 is if readline support is not enabled (default) when building the client.
If you install Bacula on more than one system, and they are identical, you can simply transfer the source tree to that other system and do a make install. However, if there are differences in the libraries or OS versions, or you wish to install on a different OS, you should start from the original compress tar file. If you do transfer the source tree, and you have previously done a ./configure command, you must do:
make distclean
prior to doing your new ./configure. This is because the GNU autoconf tools cache the configuration, and if you re-use a configuration for a Linux machine on a Solaris, you can be sure your build will fail. To avoid this, as mentioned above, either start from the tar file, or do a make distclean.
In general, you will probably want to supply a more complicated configure statement to ensure that the modules you want are built and that everything is placed into the correct directories.
For example, on Fedora, Red Hat, or SuSE one could use the following:
CFLAGS="-g -Wall" \ ./configure \ --sbindir=/opt/bacula/bin \ --sysconfdir=/opt/bacula/etc \ --with-pid-dir=/var/run \ --with-subsys-dir=/var/run \ --with-mysql \ --with-working-dir=/opt/bacula/working \ --with-dump-email=$USER
The advantage of using the above configuration to start is that everything will be put into a single directory, which you can later delete once you have run the examples in the next chapter and learned how Bacula works. In addition, the above can be installed and run as non-root.
For the developer's convenience, I have added a defaultconfig script to the examples directory. This script contains the statements that you would normally use, and each developer/user may modify them to suit his needs. You should find additional useful examples in this directory as well.
The -enable-conio or -enable-readline options are useful because they provide a command line history, editing capability for the Console program and tab completion on various option. If you have included either option in the build, either the termcap or the ncurses package will be needed to link. On most systems, including Red Hat and SuSE, you should include the ncurses package. If Bacula's configure process finds the ncurses libraries, it will use those rather than the termcap library. On some systems, such as SuSE, the termcap library is not in the standard library directory. As a consequence, the option may be disabled or you may get an error message such as:
/usr/lib/gcc-lib/i586-suse-linux/3.3.1/.../ld: cannot find -ltermcap collect2: ld returned 1 exit status
while building the Bacula Console. In that case, you will need to set the LDFLAGS environment variable prior to building.
export LDFLAGS="-L/usr/lib/termcap"
The same library requirements apply if you wish to use the readline subroutines for command line editing, history and tab completion or if you are using a MySQL library that requires encryption. If you need encryption, you can either export the appropriate additional library options as shown above or, alternatively, you can include them directly on the ./configure line as in:
LDFLAGS="-lssl -lcyrpto" \ ./configure <your-options>
On some systems such as Mandriva, readline tends to gobble up prompts, which makes it totally useless. If this happens to you, use the disable option, or if you are using version 1.33 and above try using -enable-conio to use a built-in readline replacement. You will still need either the termcap or the ncurses library, but it is unlikely that the conio package will gobble up prompts.
readline is no longer supported after version 1.34. The code within Bacula remains, so it should be usable, and if users submit patches for it, we will be happy to apply them. However, due to the fact that each version of readline seems to be incompatible with previous versions, and that there are significant differences between systems, we can no longer afford to support it.
There are a number of options and important considerations given below that you can skip for the moment if you have not had any problems building Bacula with a simplified configuration as shown above.
If the ./configure process is unable to find specific libraries (e.g. libintl, you should ensure that the appropriate package is installed on your system. Alternatively, if the package is installed in a non-standard location (as far as Bacula is concerned), then there is generally an option listed below (or listed with ./configure -help that will permit you to specify the directory that should be searched. In other cases, there are options that will permit you to disable to feature (e.g. -disable-nls).
If you want to dive right into it, we recommend you skip to the next chapter, and run the example program. It will teach you a lot about Bacula and as an example can be installed into a single directory (for easy removal) and run as non-root. If you have any problems or when you want to do a real installation, come back to this chapter and read the details presented below.
The following command line options are available for configure to customize your installation.
By default, Bacula will install the Unix man pages in /usr/share/man/man1 and /usr/share/man/man8. If you wish the man page to be installed in a different location, use this option to specify the path. Note, the main HTML and PDF Bacula documents are in a separate tar file that is not part of the source distribution.
Qt4 is available on OpenSUSE 10.2, CentOS 5, Fedora, and Debian. If it is not available on your system, you can download the depkgs-qt package from the Bacula Source Forge download area and build it. See the INSTALL file in that package for more details. In particular to use the Qt4 built by depkgs-qt you must source the file qt4-paths.
SQLite2 is not thread safe. Batch insert cannot be enabled when using SQLite2
On most systems, MySQL, PostgreSQL and SQLite3 are thread safe.
To verify that your PostgreSQL is thread safe, you can try this (change the path to point to your particular installed libpq.a; these commands were issued on FreeBSD 6.2):
$ nm /usr/local/lib/libpq.a | grep PQputCopyData 00001b08 T PQputCopyData $ nm /usr/local/lib/libpq.a | grep mutex U pthread_mutex_lock U pthread_mutex_unlock U pthread_mutex_init U pthread_mutex_lock U pthread_mutex_unlock
The above example shows a libpq that contains the required function PQputCopyData and is thread enabled (i.e. the pthread_mutex* entries). If you do not see PQputCopyData, your version of PostgreSQL is too old to allow batch insert. If you do not see the mutex entries, then thread support has not been enabled. Our tests indicate you usually need to change the configuration options and recompile/reinstall the PostgreSQL client software to get thread support.
Bacula always links to the thread safe MySQL libraries.
Running with Batch Insert turned on is recommended because it can significantly improve attribute insertion times. However, it does put a significantly larger part of the work on your SQL engine, so you may need to pay more attention to tuning it. In particular, Batch Insert can require large temporary table space, and consequently, the default location (often /tmp) may run out of space causing errors. For MySQL, the location is set in my.conf with “tmpdir”. You may also want to increase the memory available to your SQL engine to further improve performance during Batch Inserts.
When linking a static binary, the linker needs the static versions of all the libraries that are used, so frequently users will experience linking errors when this option is used. The first thing to do is to make sure you have the static glibc library installed on your system. The second thing to do is the make sure you do not specify -openssl or other options on your ./configure statement as these options require additional libraries. You may be able to enable those options, but you will need to load additional static libraries.
When linking a static binary, the linker needs the static versions of all the libraries that are used, so frequently users will experience linking errors when this option is used. The first thing to do is to make sure you have the static glibc library installed on your system. The second thing to do is the make sure you do not specify -openssl or other options on your ./configure statement as these options require additional libraries. You may be able to enable those options, but you will need to load additional static libraries.
When linking a static binary, the linker needs the static versions of all the libraries that are used, so frequently users will experience linking errors when this option is used. The first thing to do is to make sure you have the static glibc library installed on your system. The second thing to do is the make sure you do not specify -openssl or other options on your ./configure statement as these options require additional libraries. You may be able to enable those options, but you will need to load additional static libraries.
When linking a static binary, the linker needs the static versions of all the libraries that are used, so frequently users will experience linking errors when this option is used. The first thing to do is to make sure you have the static glibc library installed on your system. The second thing to do is the make sure you do not specify -openssl or other options on your ./configure statement as these options require additional libraries. You may be able to enable those options, but you will need to load additional static libraries.
When linking a static binary, the linker needs the static versions of all the libraries that are used, so frequently users will experience linking errors when this option is used. The first thing to do is to make sure you have the static glibc library installed on your system. The second thing to do is the make sure you do not specify -openssl or other options on your ./configure statement as these options require additional libraries. You may be able to enable those options, but you will need to load additional static libraries.
See the note below under the -with-postgresql item.
Note, for Bacula to be configured properly, you must specify one of the four database options supported. That is: -with-sqlite, -with-sqlite3, -with-mysql, or -with-postgresql, otherwise the ./configure will fail.
If you get errors linking, you need to load the development libraries, or you need to disable SSL by setting without-openssl.
For more information on configuring and testing TCP wrappers, please see the Configuring and Testing TCP Wrappers section in the Security Chapter.
On SuSE, the libwrappers libraries needed to link Bacula are contained in the tcpd-devel package. On Red Hat, the package is named tcp_wrappers.
This option is designed primarily for use in regression testing. Most users can safely ignore this option.
Note, many other options are presented when you do a ./configure -help, but they are not implemented.
For most systems, we recommend starting with the following options:
./configure \ --enable-smartalloc \ --sbindir=/opt/bacula/bin \ --sysconfdir=/opt/bacula/etc \ --with-pid-dir=/opt/bacula/working \ --with-subsys-dir=/opt/bacula/working \ --with-working-dir=/opt/bacula/working
If you want to install Bacula in an installation directory rather than run it out of the build directory (as developers will do most of the time), you should also include the -sbindir and -sysconfdir options with appropriate paths. Neither are necessary if you do not use "make install" as is the case for most development work. The install process will create the sbindir and sysconfdir if they do not exist, but it will not automatically create the pid-dir, subsys-dir, or working-dir, so you must ensure that they exist before running Bacula for the first time.
Using SQLite:
CFLAGS="-g -Wall" \btool{./configure} \ --sbindir=/opt/bacula/bin \ --sysconfdir=/opt/bacula/etc \ --enable-smartalloc \ --with-sqlite=$HOME/bacula/depkgs/sqlite \ --with-working-dir=/opt/bacula/working \ --with-pid-dir=/opt/bacula/working \ --with-subsys-dir=/opt/bacula/working \ --enable-bat \ --enable-\texttt{readline}
or
CFLAGS="-g -Wall" \btool{./configure} \ --sbindir=/opt/bacula/bin \ --sysconfdir=/opt/bacula/etc \ --enable-smartalloc \ --with-mysql \ --with-working-dir=/opt/bacula/working --with-pid-dir=/opt/bacula/working \ --with-subsys-dir=/opt/bacula/working --enable-\texttt{readline}
or finally, a completely traditional Red Hat Linux install, which we do not recommend, because it make it harder to backup Bacula for disaster recovery purposes:
CFLAGS="-g -Wall" \btool{./configure} \ --sbindir=/usr/sbin \ --sysconfdir=/etc/bacula \ --with-scriptdir=/etc/bacula \ --enable-smartalloc \ --enable-bat \ --with-mysql \ --with-working-dir=/var/bacula \ --with-pid-dir=/var/run \ --enable-\texttt{readline}
Note, Bacula assumes that /var/bacula, /var/run, and /var/lock/subsys exist so it will not automatically create them during the install process.
To build Bacula from source, you will need the following installed on your system (they are not by default): libiconv, gcc 3.3.2, stdc++, libgcc (for stdc++ and gcc_s libraries), make 3.8 or later.
You will probably also need to: Add /usr/local/bin to PATH and Add /usr/ccs/bin to PATH for ar.
It is possible to build Bacula on Solaris with the Solaris compiler, but we recommend using GNU C++ if possible.
A typical configuration command might look like:
#!/bin/sh CFLAGS="-g" \btool{./configure} \ --sbindir=/opt/bacula/bin \ --sysconfdir=/opt/bacula/etc \ --with-mysql \ --enable-smartalloc \ --with-pid-dir=/opt/bacula/working \ --with-subsys-dir=/opt/bacula/working \ --with-working-dir=/opt/bacula/working
Note, you may need to install the following packages to build Bacula from source:
SUNWbinutils, SUNWarc, SUNWhea, SUNWGcc, SUNWGnutls SUNWGnutls-devel SUNWGmake SUNWgccruntime SUNWlibgcrypt SUNWzlib SUNWzlibs SUNW\texttt{readline} SUNWbinutilsS SUNWGmakeS SUNWlibm export PATH=/usr/bin::/usr/ccs/bin:/etc:/usr/openwin/bin:/usr/local/bin:/usr/sfw/bin:/opt/sfw/bin:/usr/ucb:/usr/sbin
If you have installed special software not normally in the Solaris libraries, such as OpenSSL, or the packages shown above, then you may need to add /usr/sfw/lib to the library search path. Probably the simplest way to do so is to run:
setenv LDFLAGS "-L/usr/sfw/lib -R/usr/sfw/lib"
Prior to running the ./configure command.
Alternatively, you can set the LD_LIBARY_PATH and/or the LD_RUN_PATH environment variables appropriately.
It is also possible to use the crle program to set the library search path. However, this should be used with caution.
Please see: The FreeBSD Diary for a detailed description on how to make Bacula work on your system. In addition, users of FreeBSD prior to 4.9-STABLE dated Mon Dec 29 15:18:01 2003 UTC who plan to use tape devices, please see the Tape Testing section of Bacula Enterprise Problems Resolution guide for important information on how to configure your tape drive for compatibility with Bacula.
If you are using Bacula with MySQL, you should take care to compile MySQL with FreeBSD native threads rather than LinuxThreads, since Bacula is normally built with FreeBSD native threads rather than LinuxTreads. Mixing the two will probably not work.
To install the binary Win32 version of the File daemon please see the Win32 Installation Chapter in this document.
The following script could be used if you want to put everything in a single directory (except for the working directory):
#!/bin/sh CFLAGS="-g -Wall" \ \btool{./configure} \ --sbindir=$HOME/bacula/bin \ --sysconfdir=$HOME/bacula/bin \ --mandir=$HOME/bacula/bin \ --enable-smartalloc \ --enable-bat \ --with-pid-dir=$HOME/bacula/bin/working \ --with-subsys-dir=$HOME/bacula/bin/working \ --with-mysql \ --with-working-dir=$HOME/bacula/bin/working \ --with-dump-email=$USER@your-site.com \ --with-job-email=$USER@your-site.com \ --with-smtp-host=mail.your-site.com exit 0
You may also want to put the following entries in your /etc/services file as it will make viewing the connections made by Bacula easier to recognize (i.e. netstat -a):
bacula-dir 9101/tcp bacula-fd 9102/tcp bacula-sd 9103/tcp
Before setting up your configuration files, you will want to install Bacula in its final location. Simply enter:
make install
If you have previously installed Bacula, the old binaries will be overwritten, but the old configuration files will remain unchanged, and the "new" configuration files will be appended with a .new. Generally if you have previously installed and run Bacula you will want to discard or ignore the configuration files with the appended .new.
If you run the Director and the Storage daemon on one machine and you wish to back up another machine, you must have a copy of the File daemon for that machine. If the machine and the Operating System are identical, you can simply copy the Bacula File daemon binary file bacula-fd as well as its configuration file bacula-fd.conf then modify the name and password in the conf file to be unique. Be sure to make corresponding additions to the Director's configuration file (bacula-dir.conf).
If the architecture or the OS level are different, you will need to build a File daemon on the Client machine. To do so, you can use the same ./configure command as you did for your main program, starting either from a fresh copy of the source tree, or using make distclean before the ./configure.
Since the File daemon does not access the Catalog database, you can remove the -with-mysql or -with-sqlite options, then add -enable-client-only. This will compile only the necessary libraries and the client programs and thus avoids the necessity of installing one or another of those database programs to build the File daemon. With the above option, you simply enter make and just the client will be built.
If you wish the daemons to be automatically started and stopped when your system is booted (a good idea), one more step is necessary. First, the ./configure process must recognize your system - that is it must be a supported platform and not unknown, then you must install the platform dependent files by doing:
(become root) make install-autostart
Please note, that the auto-start feature is implemented only on systems that we officially support (currently, FreeBSD, Red Hat/Fedora Linux, and Solaris), and has only been fully tested on Fedora Linux.
The make install-autostart will cause the appropriate startup scripts to be installed with the necessary symbolic links. On Red Hat/Fedora Linux systems, these scripts reside in /etc/rc.d/init.d/bacula-dir, /etc/rc.d/init.d/bacula-fd, and /etc/rc.d/init.d/bacula-sd. However the exact location depends on what operating system you are using.
If you only wish to install the File daemon, you may do so with:
make install-autostart-fd
To simply build a new executable in any directory, enter:
make
To clean out all the objects and binaries (including the files named 1, 2, or 3, which are development temporary files), enter:
make clean
To really clean out everything for distribution, enter:
make distclean
note, this cleans out the Makefiles and is normally done from the top level directory to prepare for distribution of the source. To recover from this state, you must redo the ./configure in the top level directory, since all the Makefiles will be deleted.
To add a new file in a subdirectory, edit the Makefile.in in that directory, then simply do a make. In most cases, the make will rebuild the Makefile from the new Makefile.in. In some case, you may need to issue the make a second time. In extreme cases, cd to the top level directory and enter: make Makefiles.
To add dependencies:
make depend
The make depend appends the header file dependencies for each of the object files to Makefile and Makefile.in. This command should be done in each directory where you change the dependencies. Normally, it only needs to be run when you add or delete source or header files. make depend is normally automatically invoked during the configuration process.
To install:
make install
This not normally done if you are developing Bacula, but is used if you are going to run it to back up your system.
After doing a make install the following files will be installed on your system (more or less). The exact files and location (directory) for each file depends on your ./configure command (e.g. if you are using SQLite instead of MySQL, some of the files will be different).
NOTE: it is quite probable that this list is out of date. But it is a starting point.
bacula bacula-dir bacula-dir.conf bacula-fd bacula-fd.conf bacula-sd bacula-sd.conf bacula-tray-monitor tray-monitor.conf bextract bls bscan btape btraceback btraceback.gdb bconsole bconsole.conf create_mysql_database dbcheck delete_catalog_backup drop_bacula_tables drop_mysql_tables make_bacula_tables make_catalog_backup make_mysql_tables mtx-changer query.sql bsmtp startmysql stopmysql bwx-console bwx-console.conf 9 man pages
The Tray Monitor is already installed if you used the -enable-tray-monitor configure option and ran make install.
As you don't run your graphical environment as root (if you do, you should change that bad habit), don't forget to allow your user to read tray-monitor.conf, and to execute bacula-tray-monitor (this is not a security issue).
Then log into your graphical environment (KDE, GNOME or something else), run bacula-tray-monitor as your user, and see if a cassette icon appears somewhere on the screen, usually on the task bar. If it doesn't, follow the instructions below related to your environment or window manager.
System tray, or notification area if you use the GNOME terminology, has been supported in GNOME since version 2.2. To activate it, right-click on one of your panels, open the menu Add to this Panel, then Utility and finally click on Notification Area.
System tray has been supported in KDE since version 3.1. To activate it, right-click on one of your panels, open the menu Add, then Applet and finally click on System Tray.
Read the documentation to know if the Freedesktop system tray standard is supported by your window manager, and if applicable, how to activate it.
See the chapter Configuring Bacula in this manual for instructions on how to set Bacula configuration files.