Difference between revisions of "ManualInstallation"

From Request Tracker Wiki
Jump to navigation Jump to search
(Adding categories)
 
(update LogToSTDERR explanation to match the rest of the deployment strategy)
 
(66 intermediate revisions by 14 users not shown)
Line 1: Line 1:
Prev: [[ManualRequirements]] --- Up: [[UserManual]] --- Next: [[ManualApacheConfig]]
Prev: [[ManualRequirements]] Up: [[UserManual]] Next: [[ManualApacheConfig]]


----
This guide walks you through installing RT from source on a modern, popular Linux distro. Specifically, that means a distribution based on Debian or Red Hat that’s been released since around 2020.


= GENERAL INSTALLATION =
This guide assumes:


== 1. Unpack ==
* You can install packages generally available in Debian/Ubuntu or Red Hat/Fedora/CentOS.
* You want to install RT, and all of its Perl dependencies, from source to get the latest versions. (This is a trade-off. It means the boundaries of your install will be very clear, but you won’t get security updates for RT or Perl modules from your distribution.)
* You are willing to install a couple of extra tools to manage the RT installation similarly to how you would in other packaging systems (like PyPI, npm, etc.).
* You are willing to do a relatively maximal install of RT, enabling all the options during installation and then setting what you need in the configuration. (You could save a little space and time by being pickier about your options, but then that complicates the guide and makes it harder to turn those options on later if you want.)
* You are using a regular user account on the Linux system that can get superuser privileges with sudo.


== 2. Run the configure script ==
== Install the base dependencies ==


From within the temporary directory you just created, run the "configure" script. You'll want to read this entire section, and decide on the option switches which you will want to specify to configure, before you actually run it. Having multiple console windows available during (this or any other) install is '''exceptionally''' helpful.
These are required by RT, either to run or to install the dependencies.


To see a list of configuration options, use the "help" flag, which will show you a list of options:
=== Debian/Ubuntu ===


./configure --help
<pre>sudo apt install autoconf build-essential cpanminus curl libexpat-dev libgd-dev libssl-dev libz-dev gnupg graphviz multiwatch openssl perl w3m</pre>


We've also listed the options here. Defaults are specified in brackets.
=== Red Hat Enterprise Linux ===


Configuration:
These instructions are for RHEL specifically. For RHEL-derived distributions like CentOS and Rocky, go to the next section.
  -h, --help              display this help and exit
  --help=short        display options specific to this package
  --help=recursive    display the short help of all the included packages
  -V, --version          display version information and exit
  -q, --quiet, --silent  do not print `checking...' messages
  --cache-file=FILE  cache test results in FILE [disabled]
  -C, --config-cache      alias for `--cache-file=config.cache'
  -n, --no-create        do not create output files
  --srcdir=DIR        find the sources in DIR [configure dir or `..']


=== Installation directory ===
<pre>MAJDISTVER="$(. /etc/os-release && echo "${VERSION_ID%%.*}")"
sudo subscription-manager repos --enable "codeready-builder-for-rhel-$MAJDISTVER-$(arch)-rpms"
sudo dnf install "https://dl.fedoraproject.org/pub/epel/epel-release-latest-$MAJDISTVER.noarch.rpm"
sudo dnf install patch tar which gcc gcc-c++ perl-core perl-App-cpanminus graphviz expat-devel gd-devel multiwatch openssl openssl-devel w3m
sudo sed -i~ '/^SELINUX=/ c SELINUX=disabled' /etc/selinux/config
sudo setenforce 0</pre>


By default, `make install' will install all the files in `/opt/rt3/bin', `/opt/rt3/lib' etc. You can specify an installation prefix other than `/opt/rt3' using `--prefix', for instance `--prefix=$HOME'.
(Turning off SELinux enforcement is required on Red Hat-based distributions because, as of March 2022, nobody has written a policy for RT.)


--prefix=PREFIX        install architecture-independent files in PREFIX
=== RHEL Community Distributions: Fedora/CentOS/Rocky ===
                          [/opt/rt3]
--exec-prefix=EPREFIX  install architecture-dependent files in EPREFIX
                          [PREFIX]


=== Installation directory fine-tuning ===
<pre>sudo dnf install epel-release
sudo dnf install patch tar which gcc gcc-c++ perl-core perl-App-cpanminus graphviz expat-devel gd-devel multiwatch openssl openssl-devel w3m
sudo sed -i~ '/^SELINUX=/ c SELINUX=disabled' /etc/selinux/config
sudo setenforce 0</pre>


Here are some more configure switches you can use to fine tune exactly where the install process should put its files. If you don't know why or to where you might change these, just accept the default.
(Turning off SELinux enforcement is required on Red Hat-based distributions because, as of March 2022, nobody has written a policy for RT.)


--bindir=DIR          user executables [EPREFIX/bin]
== Install a database ==
--sbindir=DIR          system admin executables [EPREFIX/sbin]
--libexecdir=DIR      program executables [EPREFIX/libexec]
--datadir=DIR          read-only architecture-independent data [PREFIX/share]
--sysconfdir=DIR      read-only single-machine data [PREFIX/etc]
--sharedstatedir=DIR  modifiable architecture-independent data [PREFIX/com]
--localstatedir=DIR    modifiable single-machine data [PREFIX/var]
--libdir=DIR          object code libraries [EPREFIX/lib]
--includedir=DIR      C header files [PREFIX/include]
--oldincludedir=DIR    C header files for non-gcc [/usr/include]
--infodir=DIR          info documentation [PREFIX/info]
--mandir=DIR          man documentation [PREFIX/man]


Optional Features:
You need access to a database server. It can be remote, or you can install a database server alongside RT. RT supports a few different databases, but the best supported options are PostgreSQL and MariaDB.


--disable-FEATURE      do not include FEATURE (same as --enable-FEATURE=no)
=== Installing and configuring the PostgreSQL server ===
--enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
--enable-layout=LAYOUT  Use a specific directory layout (Default: RT3) Optional Packages:
--with-PACKAGE[=ARG]    use PACKAGE [ARG=yes]
--without-PACKAGE      do not use PACKAGE (same as --with-PACKAGE=no)
--with-rt-group=GROUP  group to own all files (default: rt)
--with-bin-owner=OWNER  user that will own rt executable files (default root)
--with-libs-owner=OWNER user that will own RT libraries (default root)
--with-libs-group=GROUP group that will own rt binaries (default bin)
--with-db-type=TYPE    sort of database RT will use (default: mysql; others are
                        Oracle, Informix, Pg, SQLite)
--with-db-host=HOSTNAME FQDN of database server (default: localhost)
--with-db-port=PORT    port on which the database listens on
--with-db-rt-host=HOSTNAME
                        FQDN of database server (default: localhost)
--with-db-dba=DBA      name of database administrator (default: root)
--with-db-database=DBNAME
                        name of the database to use (default: rt3)
--with-db-rt-user=DBUSER
                        name of database user (default: rt_user)
--with-db-rt-pass=PASSWORD
                        password for database user (default: rt_pass)
--with-web-user=USER    user the web server runs as (default: www)
--with-web-group=GROUP  group the web server runs as (default: www)
--disable-option-checking  ignore unrecognized --enable/--with options
--disable-FEATURE      do not include FEATURE (same as --enable-FEATURE=no)
--enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
--enable-layout=LAYOUT  Use a specific directory layout (Default: relative)
--enable-devel-mode    Turn on development aids that might hurt you in production
--enable-graphviz      Turns on support for RT's GraphViz dependency charts
--enable-gd            Turns on support for RT's GD pie and bar charts
--enable-gpg            Turns on GNU Privacy Guard (GPG) support


Some influential environment variables:
If you want to install a fresh PostgreSQL database server alongside RT:


PERL        Perl interpreter command
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|<pre>sudo apt install postgresql</pre>
|<pre>sudo dnf install postgresql-server</pre>
|}


Use these variables to override the choices made by 'configure' or to help it to find libraries and programs with unusual names/locations.
In order to set up RT’s database, you will need a PostgreSQL account that can create databases and roles and be authenticated with a password. If you don’t have that, you can create it by running:


Look at all the available flags and determine your RT install preferences, then
<pre>sudo createuser --createdb --createrole --login --pwprompt rt_admin</pre>
Set the password when prompted. Record this; you’ll need it later.


./configure (with the flags you want)
=== Enable password authentication in PostgreSQL ===


You might find it useful to put all your switches into a script in your build directory called <code>build</code>, so that if you need to run it more than once, or refer back to it, it's been saved somewhere. You can then do your build by merely typing
You need to consider this step whether you install the database locally, or use an existing one already running. RT supports connecting to PostgreSQL a few different ways, but authenticating with a username and password is simplest, and this guide is written based on that. Not all PostgreSQL installations allow this authentication method by default. You need to review your <code>pg_hba.conf</code> file located at:


./build
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|<pre>/etc/postgresql/VERSION/main/pg_hba.conf</pre>
|<pre>/var/lib/pgsql/VERSION/data/pg_hba.conf</pre>
|}


== 3. Dependencies ==
Replace <code>VERSION</code> with the version of your PostgreSQL database. Add these two lines '''above''' any other lines that start with <code>host</code>:


After you build RT, but before you can run it, you must satisfy RT's myriad dependencies. RT makes use of many components external to the distributed package, and these must all be installed before the package will run. You '''must always''' do this step on install/update or upgrade actions.
<pre>host  rt5  rt_user  all md5
host  rt5  rt_admin  all  md5</pre>


First of all, check for compliance, using the CPAN-based dependency checker (if you have never run CPAN on this machine as this user, see below to manually initialize it first, or else :-):
This configuration will let <code>rt_user</code> and <code>rt_admin</code> authorize themselves for the <code>rt5</code> database using an <code>md5</code> crypted password over a network connection (possibly using the localhost loopback network). You might be able to further restrict some of these fields for improved security, but doing so is outside the scope of this install guide. Refer to [https://www.postgresql.org/docs/current/auth-pg-hba-conf.html the pg_hba.conf documentation] for more details.


make testdeps
Save your changes and reload the database:


or (the long way)
<pre>sudo systemctl reload postgresql</pre>


perl sbin/rt-test-dependencies --with-&lt;databasename&gt; --with-&lt;web-environment&gt;
=== Installing the PostgreSQL client libraries ===


Instead of &lt;databasename&gt;, type either mysql, postgres or oracle. Instead of &lt;web-environment&gt;, type either fastcgi, modperl1, or modperl2.
These are required for RT to be able to talk to any PostgreSQL server.


If there are unsatisfied dependencies (that is, if you are missing stuff), you will see complaints like "such-and-such not installed". Look carefully; these don't stand out that much. Go, ye, and fix dependencies. As root:
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|<pre>sudo apt install libpq-dev</pre>
|<pre>sudo dnf install postgresql-devel</pre>
|}


make fixdeps
Once this is done you can skip ahead to installing RT.


or
=== Installing and configuring the MariaDB server ===


perl sbin/rt-test-dependencies \
If you want to install a fresh MariaDB database server alongside RT:
--with-&lt;databasename&gt; --with-&lt;web-environment&gt; --install


'''NOTE:''' If you have having difficulties with 'make fixdeps' due to errors like:
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|<pre>sudo apt install mariadb-server</pre>
|<pre>sudo dnf install mariadb-server</pre>
|}


<nowiki>==&gt; Couldn't mkdir '/root/.cpan/build/tmp': File exists          &lt;==
In order to set up RT’s database, you will need a MySQL superuser account. To stay consistent with PostgreSQL, I suggest setting a password for it. You can do that by running:
</nowiki>


Set the following environment variable to tell fixdeps to invoke CPAN as a command line tool:
<pre>sudo mysql mysql
MariaDB [mysql]&gt; GRANT ALL PRIVILEGES ON rt5.* TO rt_admin@localhost IDENTIFIED BY 'YourPassphraseHere' WITH GRANT OPTION;</pre>
Record your passphrase; you’ll need it later.


RT_FIX_DEPS_CMD='/usr/bin/perl -MCPAN -e"install %s"'
=== Adjust MariaDB’s max_allowed_packet setting ===


Make sure to export this variable with
You need to consider this step whether you install the database locally, or use an existing one already running. [https://mariadb.com/kb/en/server-system-variables/#max_allowed_packet MariaDB’s <code>max_allowed_packet</code> setting] functionally limits the size of attachments in RT. The default is 16MiB, which is too small for most installations. You can ultimately choose any setting you’re comfortable with; 64MiB here should allow most requests without being too open.


export RT_FIX_DEPS_CMD
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|<pre>echo -e '[server]\nmax_allowed_packet=64M' | sudo tee /etc/mysql/conf.d/max_allowed_packet.cnf
sudo systemctl reload mariadb</pre>
|<pre>echo -e '[server]\nmax_allowed_packet=64M' | sudo tee /etc/my.cnf.d/max_allowed_packet.cnf
sudo systemctl reload mariadb</pre>
|}


'''NOTE:''' if you've never run CPAN on your machine as the user whom you are building RT as, run it first or this part of the install will not be fun.
=== Installing the MariaDB client libraries ===


perl -MCPAN -e shell
These are required for RT to be able to talk to any MariaDB server.


You'll be walked through the configuration process. When you're done:
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|<pre>sudo apt install libmariadb-dev libmariadb-dev-compat</pre>
|On latest versions:
<pre>sudo dnf install mariadb-connector-c-devel</pre>
If you're on an older version that doesn't have that package:
<pre>sudo dnf install mariadb-devel</pre>
|}


quit
== Install RT ==


'''NOTE:''' if things are still messed up, you are logged in as root, right?
=== Create RT system accounts ===


Alternately, you could install missing items by hand. For example:
This account will be used throughout the install process to control access to sensitive files. Run:


perl -MCPAN -e 'install DBD::mysql'
<pre>sudo groupadd --system rt
sudo useradd --system --home-dir=/opt/rt5/var --gid=rt rt</pre>


would install DBD::mysql.
=== Get and unpack the RT source code ===


'''NOTE:''' Some modules may require environment variables defined, for example <code>Apache::Request</code>.
Download the latest source code using the link on the [https://bestpractical.com/download-page RT download page], extract it using <code>tar -xf</code>, and <code>cd</code> into the source code directory to run the rest of the commands in this section. For example:


'''NOTE:''' If you are using FastCGI, you will need to make sure that the FCGI module is installed. If not, run
<pre>curl -O https://download.bestpractical.com/pub/rt/release/rt-5.0.3.tar.gz
tar -xf rt-5.0.3.tar.gz
cd rt-5.0.3</pre>


perl -MCPAN -e 'install FCGI'
=== Pre-configure RT ===


or something equivalent.
This command will detect some information about your system in order to install RT properly, and decide which set of dependencies to install. Here’s what the different parts of our command are doing:


Re-check to make sure everything was installed properly:
* <code>PERL=…</code> configures RT to run under Perl with some options that load its dependencies from a dedicated directory. The <code>/opt/rt5</code> part of this string should match your <code>prefix</code> setting (see next item). If you know you want to use a specific Perl installed on your system, you can specify its full path instead of the plain <code>perl</code> here.
* <code>--prefix=/opt/rt5</code> sets the directory where RT will install all of its libraries, tools, and supporting files. You can choose another path if you like.
* <code>--with-db-type=TYPE</code> - Replace <code>TYPE</code> with <code>Pg</code> for PostgreSQL, or <code>mysql</code> for MariaDB.
* <code>--with-web-user=rt --with-web-group=rt</code> installs RT using the dedicated accounts we created earlier.
* The rest of the options tell RT to install additional dependencies for optional features.


make testdeps
Make sure you have <code>cd</code>ed into the RT source directory, and run:


or
<pre>PERL="/usr/bin/env -S perl -I/opt/rt5/local/lib/perl5" ./configure --prefix=/opt/rt5 --with-db-type=TYPE --with-web-user=rt --with-web-group=rt --with-attachment-store=disk --enable-externalauth --enable-gd --enable-graphviz --enable-gpg --enable-smime</pre>
For more background, [https://docs.bestpractical.com/rt/latest/configure.html refer to the RT configure options documentation].


perl sbin/rt-test-dependencies --with-&lt;databasename&gt; --with-&lt;web-environment&gt;
=== Install RT and its Perl dependencies ===


'''NOTE:''' Some versions of the <code>MIME::Tools</code> package will not install cleanly. Specifically, <code>MIME::Tools</code> version 5.427 is missing a dependency that will not be installed automatically by the CPAN shell. It requires <code>MIME::Base64</code> to be installed first. Otherwise, <code>MIME::Entity</code> will fail during testing and terminate the installation. Install <code>MIME::Base64</code> explicitly, either manually or using:
This command will download, build, and install all of the Perl modules necessary to run RT with the configuration you set above. Here’s what the different parts of the command are doing:


<nowiki># perl -MCPAN -e "install 'MIME::Base64'"
* <code>make dirs</code> creates RT's directory structure, so we can install dependencies inside it.
* <code>make fixdeps</code> actually installs those dependencies.
</nowiki>
* <code>RT_FIX_DEPS_CMD=…</code> tells RT to use cpanminus to install dependencies (instead of the older, default <code>cpan</code> command) inside RT's directory structure.
* <code>make install</code> installs all of RT’s files under <code>/opt/rt5</code> (or the prefix directory you set earlier). It will only run if <code>fixdeps</code> succeeded.


Thereafter, installing dependencies worked without further problems for me.
Make sure you have <code>cd</code>ed into the RT source directory, and run:


- Ed Eaglehouse
<pre>sudo make dirs
make fixdeps RT_FIX_DEPS_CMD="cpanm --sudo --local-lib-contained=/opt/rt5/local"
sudo make install</pre>
If it works, the command will eventually output a message that says “Congratulations. RT is now installed.” followed by instructions about configuring and setting up the database. We’ll do that next.


'''NOTE:''' graphviz dependencies can be met with packages from here: http://www.graphviz.org/Download..php. On [[CentOS]] 5.4, adding the repository provided by graphviz.org and doing
=== Configure RT ===


yum install graphviz graphviz-gd graphviz-devel
[https://docs.bestpractical.com/rt/latest/RT_Config.html RT has many configuration options.] You can put configuration options in the file <code>/opt/rt5/etc/RT_SiteConfig.pm</code>, or in individual files under <code>/opt/rt5/etc/RT_SiteConfig.d/</code>. Use an editor to save all the text below to <code>/opt/rt5/etc/RT_SiteConfig.pm</code> (you can just overwrite the existing file, or add this to the bottom of what’s there) and then fill in settings for your site everywhere the text <code>EDIT WITH</code> appears.


cleared up all related requirements. Installing additional packages via
<pre>
# Single-quote all values EXCEPT the special value `undef`
# that turns off a setting.


yum install gd gd-devel
# rtname appears in ticket email subjects. It needs to be globally unique,
# so use your organization's domain name.
Set($rtname, 'EDIT WITH yourdomain.example.com');
# Organization is used in the database for ticket links, etc. It also needs to
# be globally unique, so use your organization's domain name.
Set($Organization, 'EDIT WITH yourdomain.example.com');
# WebDomain is domain name of the RT web server. RT uses it to construct links
# and defend against CSRFs.
Set($WebDomain, 'EDIT WITH rt.yourdomain.example.com');
# WebPort is the port where the RT web server runs. Edit the number below if
# you're not using the standard HTTPS port.
Set($WebPort, '443');
# WebPath is the path where the RT web server runs on your WebDomain.
# Edit the path below if you're using a specific path.
Set($WebPath, '/');


and
# DatabaseUser is the name of the database account RT uses to read and store
# data. 'rt_user' is the default but you can change it if you like.
# DO NOT use the 'rt_admin' superuser created in the instructions above.
Set($DatabaseUser, 'rt_user');
# DatabasePassword is the password for DatabaseUser.
Set($DatabasePassword, 'EDIT WITH SomePassphraseHere');
# DatabaseHost is the hostname of the database server RT should use.
# Change 'localhost' if it lives on a different server.
Set($DatabaseHost, 'localhost');
# DatabasePort is the port number of the database server RT should use.
# `undef` means the default for that database. Change it if you're not
# using the standard port.
Set($DatabasePort, undef);
# DatabaseName is the name of RT's database hosted on DatabaseHost.
# 'rt5' is the default but you can change it if you like.
Set($DatabaseName, 'rt5');
# DatabaseAdmin is the name of the user in the database used to perform
# major administrative tasks. Change 'rt_admin' if you're using a user
# besides the one created in this guide.
Set($DatabaseAdmin, 'rt_admin');


yum install expat expat-devel
# RT can log to syslog, stderr, and/or a dedicated file.
# Log settings are used both by the primary server and by command line
# tools like rt-crontool, rt-ldapimport, etc.
# You set all of RT's $LogTo* paramaters to a standard log level: 'debug',
# 'info', 'notice', 'warning', 'error', 'critical', 'alert', or 'emergency'.
# For a modern install, I recommend logging primarily to syslog, so it goes
# to journald where it's easy to query and automatically gets rotated.
Set($LogToSyslog, 'info');


resolved all dependency problems. Doing 'make fixdeps' after a configure stage such as
# When the RT server logs to stderr, that will go to the rt-server journal.
# Command line tools log to their own stderr. Setting this to
# 'warning' or 'error' helps ensure you get notified if RT's cron jobs
# encounter problems.
Set($LogToSTDERR, 'warning');


./configure --with-web-handler=modperl2 --enable-graphviz --enable-gd --enable-gpg
# Turn off optional features that require additional configuration.
# If you want to use these, refer to the RT_Config documentation for
# instructions on how to set them up.
Set(%GnuPG, 'Enable' => '0');
Set(%SMIME, 'Enable' => '0');


completed with all dependencies met.
# Perl expects to find this 1 at the end of the file.
1;
</pre>


== 4. rt group ==
<code>RT_SiteConfig.pm</code> is actually Perl code. RT runs the code directly to load the configuration. Any time you finish editing it, you can check that you didn’t make any syntax errors by running:


Create a new Unix group called 'rt' (or whatever you gave to the --with-rt-group option to the configure script).
<pre>perl -c /opt/rt5/etc/RT_SiteConfig.pm</pre>


== 5. For new installations only ==
=== Set up RT’s database ===


If you're upgrading within the RT 3.x series, '''skip to step 6''':
RT includes a tool to help you set up its database. By default, it connects to the database as an administrator to create the database and user that you configured in the previous step.


=== 5.1 Init DB ===
(The instructions from <code>make install</code> and RT’s README file tell you to run <code>make initialize-database</code>. That just runs <code>rt-setup-database</code> for you. Running the tool directly makes it easier to pass the options you need.)


As root and from within the rt build directory, type:
* <code>--action=init</code> tells the tool to create the user, the database, the tables inside it, and insert core data RT needs to function.
* '''If''' you are using an existing database server and the database adminstrator has already created the user account and database for RT, then you can add the <code>--skip-create</code> option.
* '''If''' you have a less common database setup, this tool has additional options to give you finer-grained control over what steps are run and how. Refer to [https://docs.bestpractical.com/rt/latest/rt-setup-database.html the full rt-setup-database documentation] to learn more about those.
* The command reads files from RT’s <code>etc/</code> directory by default, so the easiest way to run it is to <code>cd /opt/rt5</code> first, and then it will find the necessary files automatically.


make install
Run:


Now you need to modify the installed [=etc/[[RT SiteConfig|RT_SiteConfig]].pm] to specify the connections to your DBMS engine and then type:
<pre>cd /opt/rt5
sudo sbin/rt-setup-database --action=init</pre>
Enter the password for your database administrator account when prompted.


make initialize-database
=== Set up fulltext indexing ===


For some database backends ([[MySQL]] at least) it is not able to create the database user. So you have to create that user beforehand and you have to give it rights for the database.
Fulltext indexing speeds up searches for ticket content, which makes RT a lot nicer to use.


For [[MySQL]] with the default [=$DatabaseUser], [=$DatabaseName] and [=$DatabasePassword], do this SQL command (as root database user):
* <code>--noask</code> uses the default names for the index, which will be fine for a new install and simplifies the setup.


GRANT ALL PRIVILEGES ON rt3.* TO 'rt_user'@'localhost' IDENTIFIED BY 'rt_pass'
Run:


Replace the values 'rt3', 'rt_user', 'localhost' and 'rt_pass' by whatever is appropriate for you.
<pre>sudo /opt/rt5/sbin/rt-setup-fulltext-index --noask</pre>
Enter the password for your database administrator account when prompted. The end of the process will output some RT configuration that looks like this:


=== 5.2 Drop DB if something goes wrong ===
<pre>### EXAMPLE OUTPUT ONLY - Don't use this directly!
Set( %FullTextSearch,
    Enable    => 1,
    Indexed    => 1,
    # Additional output from rt-setup-fulltext-index should be here.
    # The configuration varies by database type.
);</pre>
Copy the output generated when you run <code>rt-setup-fulltext-index</code> and save it to the file <code>/opt/rt5/etc/RT_SiteConfig.d/FulltextIndex.pm</code>.


If the make fails, type:
=== Set permissions ===


make dropdb
All of RT’s configuration files should be readable by the user that runs the web server, and no other users, in order to protect sensitive information like the database password. RT provides a command to set permissions appropriately according to your distribution and configuration. <code>cd</code> to the directory where you extracted the RT source code, and run:


fix whatever's broken, and start over from step "init DB" step
<pre>cd rt-5.0.3
sudo make fixperms</pre>
=== Verify the installation ===


=== 5.3 if still failing... ===
If everything has gone well, then you should be able to set a password for RT’s <code>root</code> user. You’ll use this later to log in to the web interface and continue setting up your system. Run:


If make install still fails, look in <code>/etc/httpd/httpd.conf</code> or <code>/etc/httpd/conf/commonhttpd.conf</code> (or wherever your httpd.conf is; this may vary by Unix distribution; <code>locate httpd.conf</code> may help) for a line that reads "Group &lt;something&gt;" and another that reads "User &lt;somebody&gt;"
<pre>sudo /opt/rt5/sbin/rt-passwd root</pre>
Set the password when prompted. Record this; you’ll need it later.


Go to your RT source directory and type (without the brackets, substituting the right Group and User from the file you just looked at):
== Set up RT FastCGI service ==


./configure --with-web-group=&lt;something&gt; --with-web-user=&lt;somebody&gt;
=== Create the RT service ===


Skip to step 7.
Run:


=== 5.4 initdb fails ===
<pre>sudo systemctl edit --force --full rt-server.service</pre>


The defaults were changed in [[PostgreSQL]] 8.1 and now tables are created without [[OIDs]]. One of workarounds is to add the following block at the top of <code>etc/schema.Pg</code> file:
This opens an editor to define the RT service to systemd. Use the editor to add this text to the file:


--
<pre>[Unit]
-- Enforce OID creation for 8.1+
Description=RT FCGI server
--
set default_with_oids = 't';


Once it's done "make initialize-database" will work perfectly.
[Service]
# The --forks option is the number of RT servers to run in parallel.
# 3 should be good for most initial installs. You can increase this
# number later if needed for performance.
ExecStart=/usr/bin/multiwatch --forks=3 --signal=TERM -- /opt/rt5/sbin/rt-server.fcgi
StandardInput=socket
User=rt
UMask=027


-- Added by GalaxyMaster &lt;galaxy at openwall.com&gt;
CapabilityBoundingSet=
DevicePolicy=closed
PrivateMounts=true
PrivateNetwork=false
PrivateTmp=true
PrivateUsers=true
ProtectControlGroups=true
ProtectHome=true
ProtectSystem=full
</pre>


== 6. For an Upgrade ==
Save the file and close your editor.


For upgrading within the RT 3.x series - if you are not upgrading or have already completed step 5, '''skip to step 7'''.
=== Create the RT socket ===


As root, type:
Run:


make upgrade
<pre>sudo systemctl edit --force --full rt-server.socket</pre>


(replace "make" with the local name for Make, if you need to)
This opens an editor to define the RT socket to systemd. Use the editor to add this text to the file:


This will build new executable files, config files and libraries without overwriting your RT database.
<pre>[Unit]
Description=RT FCGI server socket
Wants=network.target
After=network.target
Before=apache2.service httpd.service nginx.service


It may then instruct you to update your RT system database objects
[Install]
WantedBy=sockets.target


== 7. Configuration ==
[Socket]
# ListenStream defines the address and port where the RT FastCGI server listens.
# This is NOT the web server itself, so don't make this port 80, 443, etc.
# You may edit this if you like, but note:
# Connections are unencrypted. You should only listen on a secure network
# interface.
# The server can only accept a single socket. You cannot specify more than
# one ListenStream address.
ListenStream=[::1]:5000
Accept=no
FreeBind=yes
</pre>


Edit [=etc/[[RT SiteConfig|RT_SiteConfig]].pm] in your RT installation directory, by specifying any values you need to change from the defaults as defined in RT_Config.pm. It is easiest to do this by copying RT_Config.pm to RT_SiteConfig.pm, and then uncommenting and changing anything you need to set, though perhaps this isn't quite the '''best''' approach.
Save the file and close your editor.


In many cases sensible defaults have been included. In others, you must supply a value. Some values (such as the RT log directory) will come from values you supplied in the Makefile. You'll find further explanation inline in the [[RT SiteConfig|RT_SiteConfig]].pm file. You should look at and consider changing the following entries:
=== Enable the RT socket ===


$DatabasePassword = 'rt_pass'
Start the server socket and enable it on future boots:


which is the password the [[DatabaseUser]] should use to access the database.
<pre>sudo systemctl enable --now rt-server.socket</pre>


'''NOTE:''' Some [[MySQL]] users have had trouble with passwords of longer than 8 characters; if you cannot connect, try a password of 8 characters or fewer.
== Set up RT’s web server ==


$CanonicalizeEmailAddressMatch = 'subdomain.example.com$';
Fully documenting how to set up a web server configured for your network with encryption is outside the scope of this guide. This page only highlights the configuration you need to pass web requests onto the RT FastCGI service.
$CanonicalizeEmailAddressReplace = 'example.com';


The $[[CanonicalizeEmailAddress]] variables allow you to keep incoming messages consistent, such as when a site removes the subdomain from an email address. In the example presented by the defaults, if messages from your organization sometimes come from exchange.example.com and sometimes from example.com, you'd set $[[CanonicalizeEmailAddressMatch]] to exchange.example.com and $[[CanonicalizeEmailAddressReplace]] to example.com
=== nginx ===


$SenderMustExistInExternalDatabase = undef;
Make sure you have the necessary packages installed:


If $[[SenderMustExistInExternalDatabase]] is true, RT will refuse to auto-create non-staff accounts for unknown users filing new tickets by email if you are using the "[[LookupSenderInExternalDatabase]]" option elsewhere in [[RT SiteConfig|RT_SiteConfig]].pm. Instead, an error message will be returned and RT will forward the user's message to $[[RTOwner]] as defined in [[RT SiteConfig|RT_SiteConfig]].pm. If you are not using $[[LookupSenderInExternalDatabase]], this option has no effect. If you define an [[AutoRejectRequest]] template, RT will use this template for the rejection message.
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|<pre>sudo apt install nginx</pre>
|<pre>sudo dnf install nginx</pre>
|}


$CorrespondAddress = 'RT::CorrespondAddress.not.set';
In the <code>http server</code> configuration blocks where you want to serve RT, add a configuration block like this, '''above''' any other <code>location</code> blocks:
$CommentAddress = 'RT::CommentAddress.not.set';


$[[CorrespondAddress]] and $[[CommentAddress]] are the default addresses that will be listed in both From: and Reply-To: headers of reply and comment mail, respectively, sent by RT, unless they are overridden by a queue-specific address.
<pre>
# The location path should match the WebPath in your RT site configuration.
location / {
  include /etc/nginx/fastcgi.conf;
  # SCRIPT_NAME should match RT's WebPath, without a trailing slash.
  # This means when WebPath is /, it's the empty string "".
  fastcgi_param SCRIPT_NAME "";
  # This network location should match the ListenStream in rt-server.socket.
  fastcgi_pass localhost:5000;
}
</pre>


$MailCommand = 'sendmailpipe';
If you're doing a fresh install, the default <code>http server</code> configuration location varies by distribution:


$[[MailCommand]] defines which method RT will use to try to send mail. We know sendmailpipe works fairly well. If sendmailpipe doesn't work well for you, try sendmail. Note that you should remove the '-t' from $[[SendmailArguments]] if you use sendmail rather than sendmailpipe. Also note that sendmailpipe and sendmail aren't the names of commands on your system, but instructions that tell RT what mail delivery subsystem to try.
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|Edit <code>/etc/nginx/sites-available/default</code>
|Add a new file <code>/etc/nginx/default.d/rt-server.conf</code>
|}


*Don't forget to restart the Apache webserver after doing changes in [[RT SiteConfig|RT_SiteConfig]].pm!* This is true of any change, but we mention it here since this is the configuration option you're most likely to have to experiment with.
After you’ve edited your configuration, load it by running:


$SendmailArguments = "-oi -t";
<pre>sudo systemctl reload nginx</pre>


$[[SendmailArguments]] defines what flags to pass to $Sendmail, assuming you picked sendmail or sendmailpipe as the $[[MailCommand]]. If you picked sendmailpipe, then $[[SendmailArguments]] must include the "-t" flag. The default options are good for most sendmail wrappers and workalikes.
Once this is done you can skip ahead to Verify the web interface.


$SendmailPath = "/usr/sbin/sendmail";
=== Apache httpd ===


If you selected sendmailpipe as $[[MailCommand]], you must specify the path to your sendmail executable file in $[[SendmailPath]]. If you did not select sendmailpipe this has no effect.
Make sure you have the necessary packages installed, and the proxy_fcgi module enabled:


In case of '''Exim''', the following configuration works:
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|<pre>sudo apt install apache2
sudo a2enmod proxy
sudo a2enmod proxy_fcgi</pre>
|<pre>sudo dnf install httpd</pre>
|}


Set($MailCommand , 'sendmail');
In the <code>&lt;VirtualHost&gt;</code> configuration blocks where you want to serve RT, add a configuration block like this, '''above''' any other <code>Alias</code>, <code>ProxyPass</code>, or <code>ScriptAlias</code> lines:
Set($SendmailArguments , "-bm -- &lt;your email address&gt;");
Set($SendmailPath, "/usr/sbin/exim4");
Set($NotifyActor, 1);
$Timezone = 'US/Eastern';


$Timezone is used to convert times entered by users into GMT and back again. It should be set to a timezone recognized by your local Unix box, and -- in general -- you should pick the timezone the majority of your users reside in.
<pre>
# SCRIPT_NAME should match RT's WebPath, without a trailing slash.
# This means when WebPath is /, it's the empty string "".
ProxyFCGISetEnvIf true SCRIPT_NAME ""
# The location path in the first argument should match the WebPath in your
# RT site configuration.  The network location after fcgi:// should match
# the ListenStream in rt-server.socket.
ProxyPass / fcgi://localhost:5000/
</pre>


$UseFriendlyToLine = 0;
If you're doing a fresh install, the default <code>&lt;VirtualHost&gt;</code> configuration location varies by distribution and whether or not you're using HTTPS:


RT can set a "friendly", rather than blank, To: header when sending messages to Ccs or [[AdminCcs]]. This feature does not work with Sendmail(tm)-brand sendmail. If you are using sendmail, rather than postfix, qmail, exim, or some other program, you must disable this option (by setting it to 0 rather than 1).
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|With HTTPS: <code>/etc/apache2/sites-available/default-ssl.conf</code>
With plain HTTP: <code>/etc/apache2/sites-available/000-default.conf</code>
|With HTTPS: <code>/etc/httpd/conf.d/ssl.conf</code>
With plain HTTP: Write a new file <code>/etc/httpd/conf.d/vhost_RT.conf</code>
|}


$WebPath = "";
After you’ve edited your configuration, load it by running:


A variable used to help RT construct [[URLs]] that point back to RT. If you've put RT somewhere other than at the root of your webserver, you need to define a [[WebPath]]. RT uses this in the construction of relative [[URLs]]. $[[WebPath]] requires a leading / but no trailing /
{| style="width: 100%;"
! style="width: 50%;"|Debian/Ubuntu
! style="width: 50%;"|Red Hat/Fedora/CentOS
|-
|<pre>sudo systemctl reload apache2</pre>
|<pre>sudo systemctl reload httpd</pre>
|}


Example: if your installation is at
=== Verify the web interface ===


http://www.fsck.com/rt/
You should be able to visit your web server in your browser, and be presented with RT’s login screen. You should be able to log in with username <code>root</code> and the password you set previously.


set this to "/rt".
If the web server returns a 502 Bad Gateway response, it's having trouble connecting to the RT FCGI server. Check the error logs for your web server:


$WebBaseURL = "http://not.configured:80";
{| style="width: 100%;"
! style="width: 33%;"|nginx
! style="width: 33%;"|Apache on Debian/Ubuntu
! style="width: 33%;"|httpd on Red Hat/Fedora/CentOS
|-
|<pre>/var/log/nginx/error.log</pre>
|<pre>/var/log/apache2/error.log</pre>
|<pre>/var/log/httpd/error_log</pre>
|}


A variable used to help RT construct [[URLs]] that point back to RT. $[[WebBaseURL]] is the base of the URL. it should usually include the scheme, the host, and the port if non-standard.
For all kinds of errors, you can check the logs for the RT FCGI server and its socket:


Example: "https://fsck.com" or "http://fsck.com:88"
<pre>sudo journalctl --unit rt-server.\*</pre>


$[[WebBaseURL]] doesn't need a trailing /
== Set up RT’s mail server ==


$WebURL = $WebBaseURL . $WebPath . "/";
RT both can both send and receive ticket updates via email. Unfortunately, there are too many variables to document a useful setup process here: getting this working usually requires creating DNS records, and coordinating with existing mail servers, which will be the main constraint on your setup. Instead this guide provides a brief overview of how the integration works, and where the connection points are that you likely need to work on.


A variable used to help RT construct [[URLs]] that point back to RT. [=$[[WebURL]]] is the combination of [=$[[WebBaseURL]]] and [=$[[WebPath]]]. Generally, you shouldn't change it.
=== Sending Mail ===


$WebImagesURL = $WebURL . "/NoAuth/images/";
RT only knows how to send mail by passing it off to another program on the system. It cannot connect or authenticate directly to external mail servers. In the default configuration, RT runs the standard <code>sendmail</code> command. There are [https://docs.bestpractical.com/rt/latest/RT_Config.html#Outgoing-mail configuration options to send mail through different commands] if you need.


[=$[[WebImagesURL]]] points to the base URL where RT can find its images. If you're running the [[FastCGI]] version of the RT web interface, you should make RT's [=[[WebRT]]/html/[[NoAuth]]/images] directory available on a static web server and supply that URL as [=$[[WebImagesUrl]]] -- alternately, you can tell Apache not to run it through [[FastCGI]].
The most common setup is to install and configure a proper Mail Transfer Agent (MTA) like Postfix or Exim, and then configure it to send mail to the wider Internet as you need. This works well because the MTAs are robust and well-tested; they have flexible configuration to let you send mail out by relaying to other mail servers you specify with optional authentication; and most distributions install one by default anyway. The only hard part is configuring the MTA to send mail following your site’s policies.


== THEN ==
Other software is available that provides a slimmer version of the <code>sendmail</code> command that connects to an external mail server for you, like msmtp. These programs are usually easier to configure than an MTA, but they often lose email permanently if they can’t connect to the external server at the time it’s sent. (MTAs keep email queued locally until they successfully deliver it to another server.)


Configure web server, read [[ManualApacheConfig]]
=== Receiving email ===


----
RT installs a command called <code>rt-mailgate</code> that receives an email on standard input and posts it to RT’s REST web interface, where it gets saved in the database and added to a ticket. You need to arrange for a way to send incoming email to this command.


Prev: [[ManualRequirements]] --- Up: [[UserManual]] --- Next: [[ManualApacheConfig]]
The most common setup is to have an MTA on the same box as RT receive email directly, and then set up mail aliases that call this command when mail comes in. Example <code>/etc/aliases</code> entries look like:
[[Category:RT User Manual]]
 
<pre>rt: "|/opt/rt5/bin/rt-mailgate --queue general --action correspond --url https://rt.yourdomain.example.com/"
rt-comment: "|/opt/rt5/bin/rt-mailgate --queue general --action comment --url https://rt.yourdomain.example.com/"</pre>
This works well because, again, you’re probably running an MTA anyway; and the MTA can hold and queue mail if it comes in while RT is down for any reason, giving you a buffer against downtime.
 
Another common option is to periodically run a tool that fetches mail using a protocol like IMAP, like fetchmail or getmail, and passes it on to <code>rt-mailgate</code>. This is less common because it requires setting up another tool to run, and securely storing another set of mail server credentials. But it is useful when local policy prevents the RT server from receiving email directly.
 
This is much less common, but it might help to know that <code>rt-mailgate</code> doesn’t have to run on the same system as RT itself. It just needs to be able to connect to RT’s web interface. If you don’t have any other options, you can install the RT software on a different system that receives email, and configure ''that'' system to run <code>rt-mailgate</code> and pass it on to the RT server. To do that, install the [https://metacpan.org/pod/RT::Client::CLI RT::Client::CLI] Perl package on the mail server. For example, you could install cpanminus, then run:
 
<pre>cpanm --sudo RT::Client::CLI</pre>
 
== Set up RT’s background jobs ==
 
Create a file <code>/etc/cron.d/rt</code> with the following content. You may edit all of the time fields as you see fit. Refer to the crontab(5) man page for details about their definitions.
 
<pre>
# Update the fulltext index with new ticket data
*/3 *  *  *  *  rt    /opt/rt5/sbin/rt-fulltext-indexer
# Email out dashboards that users have subscribed to
1  *  *  *  *  rt    /opt/rt5/sbin/rt-email-dashboards
# Clean old sessions from the database
50  3  *  *  *  rt    /opt/rt5/sbin/rt-clean-sessions --older 8d
# Email out weekly digests for users who have requested it
50  4  *  *  Mon rt    /opt/rt5/sbin/rt-email-digest -m weekly
# Email out daily digests for users who have requested it
50  5  *  *  *  rt    /opt/rt5/sbin/rt-email-digest -m daily
</pre>
 
== Set up RT ==
 
If you’ve gotten this far, congratulations, your RT install is really done now. You can start setting up RT with users, groups, queues, and business logic. [[Main Page|Head back to the main page]] to start exploring those topics.

Latest revision as of 14:48, 21 October 2022

Prev: ManualRequirements — Up: UserManual — Next: ManualApacheConfig

This guide walks you through installing RT from source on a modern, popular Linux distro. Specifically, that means a distribution based on Debian or Red Hat that’s been released since around 2020.

This guide assumes:

  • You can install packages generally available in Debian/Ubuntu or Red Hat/Fedora/CentOS.
  • You want to install RT, and all of its Perl dependencies, from source to get the latest versions. (This is a trade-off. It means the boundaries of your install will be very clear, but you won’t get security updates for RT or Perl modules from your distribution.)
  • You are willing to install a couple of extra tools to manage the RT installation similarly to how you would in other packaging systems (like PyPI, npm, etc.).
  • You are willing to do a relatively maximal install of RT, enabling all the options during installation and then setting what you need in the configuration. (You could save a little space and time by being pickier about your options, but then that complicates the guide and makes it harder to turn those options on later if you want.)
  • You are using a regular user account on the Linux system that can get superuser privileges with sudo.

Install the base dependencies

These are required by RT, either to run or to install the dependencies.

Debian/Ubuntu

sudo apt install autoconf build-essential cpanminus curl libexpat-dev libgd-dev libssl-dev libz-dev gnupg graphviz multiwatch openssl perl w3m

Red Hat Enterprise Linux

These instructions are for RHEL specifically. For RHEL-derived distributions like CentOS and Rocky, go to the next section.

MAJDISTVER="$(. /etc/os-release && echo "${VERSION_ID%%.*}")"
sudo subscription-manager repos --enable "codeready-builder-for-rhel-$MAJDISTVER-$(arch)-rpms"
sudo dnf install "https://dl.fedoraproject.org/pub/epel/epel-release-latest-$MAJDISTVER.noarch.rpm"
sudo dnf install patch tar which gcc gcc-c++ perl-core perl-App-cpanminus graphviz expat-devel gd-devel multiwatch openssl openssl-devel w3m
sudo sed -i~ '/^SELINUX=/ c SELINUX=disabled' /etc/selinux/config
sudo setenforce 0

(Turning off SELinux enforcement is required on Red Hat-based distributions because, as of March 2022, nobody has written a policy for RT.)

RHEL Community Distributions: Fedora/CentOS/Rocky

sudo dnf install epel-release
sudo dnf install patch tar which gcc gcc-c++ perl-core perl-App-cpanminus graphviz expat-devel gd-devel multiwatch openssl openssl-devel w3m
sudo sed -i~ '/^SELINUX=/ c SELINUX=disabled' /etc/selinux/config
sudo setenforce 0

(Turning off SELinux enforcement is required on Red Hat-based distributions because, as of March 2022, nobody has written a policy for RT.)

Install a database

You need access to a database server. It can be remote, or you can install a database server alongside RT. RT supports a few different databases, but the best supported options are PostgreSQL and MariaDB.

Installing and configuring the PostgreSQL server

If you want to install a fresh PostgreSQL database server alongside RT:

Debian/Ubuntu Red Hat/Fedora/CentOS
sudo apt install postgresql
sudo dnf install postgresql-server

In order to set up RT’s database, you will need a PostgreSQL account that can create databases and roles and be authenticated with a password. If you don’t have that, you can create it by running:

sudo createuser --createdb --createrole --login --pwprompt rt_admin

Set the password when prompted. Record this; you’ll need it later.

Enable password authentication in PostgreSQL

You need to consider this step whether you install the database locally, or use an existing one already running. RT supports connecting to PostgreSQL a few different ways, but authenticating with a username and password is simplest, and this guide is written based on that. Not all PostgreSQL installations allow this authentication method by default. You need to review your pg_hba.conf file located at:

Debian/Ubuntu Red Hat/Fedora/CentOS
/etc/postgresql/VERSION/main/pg_hba.conf
/var/lib/pgsql/VERSION/data/pg_hba.conf

Replace VERSION with the version of your PostgreSQL database. Add these two lines above any other lines that start with host:

host  rt5  rt_user   all  md5
host  rt5  rt_admin  all  md5

This configuration will let rt_user and rt_admin authorize themselves for the rt5 database using an md5 crypted password over a network connection (possibly using the localhost loopback network). You might be able to further restrict some of these fields for improved security, but doing so is outside the scope of this install guide. Refer to the pg_hba.conf documentation for more details.

Save your changes and reload the database:

sudo systemctl reload postgresql

Installing the PostgreSQL client libraries

These are required for RT to be able to talk to any PostgreSQL server.

Debian/Ubuntu Red Hat/Fedora/CentOS
sudo apt install libpq-dev
sudo dnf install postgresql-devel

Once this is done you can skip ahead to installing RT.

Installing and configuring the MariaDB server

If you want to install a fresh MariaDB database server alongside RT:

Debian/Ubuntu Red Hat/Fedora/CentOS
sudo apt install mariadb-server
sudo dnf install mariadb-server

In order to set up RT’s database, you will need a MySQL superuser account. To stay consistent with PostgreSQL, I suggest setting a password for it. You can do that by running:

sudo mysql mysql
MariaDB [mysql]> GRANT ALL PRIVILEGES ON rt5.* TO rt_admin@localhost IDENTIFIED BY 'YourPassphraseHere' WITH GRANT OPTION;

Record your passphrase; you’ll need it later.

Adjust MariaDB’s max_allowed_packet setting

You need to consider this step whether you install the database locally, or use an existing one already running. MariaDB’s max_allowed_packet setting functionally limits the size of attachments in RT. The default is 16MiB, which is too small for most installations. You can ultimately choose any setting you’re comfortable with; 64MiB here should allow most requests without being too open.

Debian/Ubuntu Red Hat/Fedora/CentOS
echo -e '[server]\nmax_allowed_packet=64M' | sudo tee /etc/mysql/conf.d/max_allowed_packet.cnf
sudo systemctl reload mariadb
echo -e '[server]\nmax_allowed_packet=64M' | sudo tee /etc/my.cnf.d/max_allowed_packet.cnf
sudo systemctl reload mariadb

Installing the MariaDB client libraries

These are required for RT to be able to talk to any MariaDB server.

Debian/Ubuntu Red Hat/Fedora/CentOS
sudo apt install libmariadb-dev libmariadb-dev-compat
On latest versions:
sudo dnf install mariadb-connector-c-devel

If you're on an older version that doesn't have that package:

sudo dnf install mariadb-devel

Install RT

Create RT system accounts

This account will be used throughout the install process to control access to sensitive files. Run:

sudo groupadd --system rt
sudo useradd --system --home-dir=/opt/rt5/var --gid=rt rt

Get and unpack the RT source code

Download the latest source code using the link on the RT download page, extract it using tar -xf, and cd into the source code directory to run the rest of the commands in this section. For example:

curl -O https://download.bestpractical.com/pub/rt/release/rt-5.0.3.tar.gz
tar -xf rt-5.0.3.tar.gz
cd rt-5.0.3

Pre-configure RT

This command will detect some information about your system in order to install RT properly, and decide which set of dependencies to install. Here’s what the different parts of our command are doing:

  • PERL=… configures RT to run under Perl with some options that load its dependencies from a dedicated directory. The /opt/rt5 part of this string should match your prefix setting (see next item). If you know you want to use a specific Perl installed on your system, you can specify its full path instead of the plain perl here.
  • --prefix=/opt/rt5 sets the directory where RT will install all of its libraries, tools, and supporting files. You can choose another path if you like.
  • --with-db-type=TYPE - Replace TYPE with Pg for PostgreSQL, or mysql for MariaDB.
  • --with-web-user=rt --with-web-group=rt installs RT using the dedicated accounts we created earlier.
  • The rest of the options tell RT to install additional dependencies for optional features.

Make sure you have cded into the RT source directory, and run:

PERL="/usr/bin/env -S perl -I/opt/rt5/local/lib/perl5" ./configure --prefix=/opt/rt5 --with-db-type=TYPE --with-web-user=rt --with-web-group=rt --with-attachment-store=disk --enable-externalauth --enable-gd --enable-graphviz --enable-gpg --enable-smime

For more background, refer to the RT configure options documentation.

Install RT and its Perl dependencies

This command will download, build, and install all of the Perl modules necessary to run RT with the configuration you set above. Here’s what the different parts of the command are doing:

  • make dirs creates RT's directory structure, so we can install dependencies inside it.
  • make fixdeps actually installs those dependencies.
  • RT_FIX_DEPS_CMD=… tells RT to use cpanminus to install dependencies (instead of the older, default cpan command) inside RT's directory structure.
  • make install installs all of RT’s files under /opt/rt5 (or the prefix directory you set earlier). It will only run if fixdeps succeeded.

Make sure you have cded into the RT source directory, and run:

sudo make dirs
make fixdeps RT_FIX_DEPS_CMD="cpanm --sudo --local-lib-contained=/opt/rt5/local"
sudo make install

If it works, the command will eventually output a message that says “Congratulations. RT is now installed.” followed by instructions about configuring and setting up the database. We’ll do that next.

Configure RT

RT has many configuration options. You can put configuration options in the file /opt/rt5/etc/RT_SiteConfig.pm, or in individual files under /opt/rt5/etc/RT_SiteConfig.d/. Use an editor to save all the text below to /opt/rt5/etc/RT_SiteConfig.pm (you can just overwrite the existing file, or add this to the bottom of what’s there) and then fill in settings for your site everywhere the text EDIT WITH appears.

# Single-quote all values EXCEPT the special value `undef`
# that turns off a setting.

# rtname appears in ticket email subjects. It needs to be globally unique,
# so use your organization's domain name.
Set($rtname, 'EDIT WITH yourdomain.example.com');
# Organization is used in the database for ticket links, etc. It also needs to
# be globally unique, so use your organization's domain name.
Set($Organization, 'EDIT WITH yourdomain.example.com');
# WebDomain is domain name of the RT web server. RT uses it to construct links
# and defend against CSRFs.
Set($WebDomain, 'EDIT WITH rt.yourdomain.example.com');
# WebPort is the port where the RT web server runs. Edit the number below if
# you're not using the standard HTTPS port.
Set($WebPort, '443');
# WebPath is the path where the RT web server runs on your WebDomain.
# Edit the path below if you're using a specific path.
Set($WebPath, '/');

# DatabaseUser is the name of the database account RT uses to read and store
# data. 'rt_user' is the default but you can change it if you like.
# DO NOT use the 'rt_admin' superuser created in the instructions above.
Set($DatabaseUser, 'rt_user');
# DatabasePassword is the password for DatabaseUser.
Set($DatabasePassword, 'EDIT WITH SomePassphraseHere');
# DatabaseHost is the hostname of the database server RT should use.
# Change 'localhost' if it lives on a different server.
Set($DatabaseHost, 'localhost');
# DatabasePort is the port number of the database server RT should use.
# `undef` means the default for that database. Change it if you're not
# using the standard port.
Set($DatabasePort, undef);
# DatabaseName is the name of RT's database hosted on DatabaseHost.
# 'rt5' is the default but you can change it if you like.
Set($DatabaseName, 'rt5');
# DatabaseAdmin is the name of the user in the database used to perform
# major administrative tasks. Change 'rt_admin' if you're using a user
# besides the one created in this guide.
Set($DatabaseAdmin, 'rt_admin');

# RT can log to syslog, stderr, and/or a dedicated file.
# Log settings are used both by the primary server and by command line
# tools like rt-crontool, rt-ldapimport, etc.
# You set all of RT's $LogTo* paramaters to a standard log level: 'debug',
# 'info', 'notice', 'warning', 'error', 'critical', 'alert', or 'emergency'.
# For a modern install, I recommend logging primarily to syslog, so it goes
# to journald where it's easy to query and automatically gets rotated.
Set($LogToSyslog, 'info');

# When the RT server logs to stderr, that will go to the rt-server journal.
# Command line tools log to their own stderr. Setting this to
# 'warning' or 'error' helps ensure you get notified if RT's cron jobs
# encounter problems.
Set($LogToSTDERR, 'warning');

# Turn off optional features that require additional configuration.
# If you want to use these, refer to the RT_Config documentation for
# instructions on how to set them up.
Set(%GnuPG, 'Enable' => '0');
Set(%SMIME, 'Enable' => '0');

# Perl expects to find this 1 at the end of the file.
1;

RT_SiteConfig.pm is actually Perl code. RT runs the code directly to load the configuration. Any time you finish editing it, you can check that you didn’t make any syntax errors by running:

perl -c /opt/rt5/etc/RT_SiteConfig.pm

Set up RT’s database

RT includes a tool to help you set up its database. By default, it connects to the database as an administrator to create the database and user that you configured in the previous step.

(The instructions from make install and RT’s README file tell you to run make initialize-database. That just runs rt-setup-database for you. Running the tool directly makes it easier to pass the options you need.)

  • --action=init tells the tool to create the user, the database, the tables inside it, and insert core data RT needs to function.
  • If you are using an existing database server and the database adminstrator has already created the user account and database for RT, then you can add the --skip-create option.
  • If you have a less common database setup, this tool has additional options to give you finer-grained control over what steps are run and how. Refer to the full rt-setup-database documentation to learn more about those.
  • The command reads files from RT’s etc/ directory by default, so the easiest way to run it is to cd /opt/rt5 first, and then it will find the necessary files automatically.

Run:

cd /opt/rt5
sudo sbin/rt-setup-database --action=init

Enter the password for your database administrator account when prompted.

Set up fulltext indexing

Fulltext indexing speeds up searches for ticket content, which makes RT a lot nicer to use.

  • --noask uses the default names for the index, which will be fine for a new install and simplifies the setup.

Run:

sudo /opt/rt5/sbin/rt-setup-fulltext-index --noask

Enter the password for your database administrator account when prompted. The end of the process will output some RT configuration that looks like this:

### EXAMPLE OUTPUT ONLY - Don't use this directly!
Set( %FullTextSearch,
    Enable     => 1,
    Indexed    => 1,
    # Additional output from rt-setup-fulltext-index should be here.
    # The configuration varies by database type.
);

Copy the output generated when you run rt-setup-fulltext-index and save it to the file /opt/rt5/etc/RT_SiteConfig.d/FulltextIndex.pm.

Set permissions

All of RT’s configuration files should be readable by the user that runs the web server, and no other users, in order to protect sensitive information like the database password. RT provides a command to set permissions appropriately according to your distribution and configuration. cd to the directory where you extracted the RT source code, and run:

cd rt-5.0.3
sudo make fixperms

Verify the installation

If everything has gone well, then you should be able to set a password for RT’s root user. You’ll use this later to log in to the web interface and continue setting up your system. Run:

sudo /opt/rt5/sbin/rt-passwd root

Set the password when prompted. Record this; you’ll need it later.

Set up RT FastCGI service

Create the RT service

Run:

sudo systemctl edit --force --full rt-server.service

This opens an editor to define the RT service to systemd. Use the editor to add this text to the file:

[Unit]
Description=RT FCGI server

[Service]
# The --forks option is the number of RT servers to run in parallel.
# 3 should be good for most initial installs. You can increase this
# number later if needed for performance.
ExecStart=/usr/bin/multiwatch --forks=3 --signal=TERM -- /opt/rt5/sbin/rt-server.fcgi
StandardInput=socket
User=rt
UMask=027

CapabilityBoundingSet=
DevicePolicy=closed
PrivateMounts=true
PrivateNetwork=false
PrivateTmp=true
PrivateUsers=true
ProtectControlGroups=true
ProtectHome=true
ProtectSystem=full

Save the file and close your editor.

Create the RT socket

Run:

sudo systemctl edit --force --full rt-server.socket

This opens an editor to define the RT socket to systemd. Use the editor to add this text to the file:

[Unit]
Description=RT FCGI server socket
Wants=network.target
After=network.target
Before=apache2.service httpd.service nginx.service

[Install]
WantedBy=sockets.target

[Socket]
# ListenStream defines the address and port where the RT FastCGI server listens.
# This is NOT the web server itself, so don't make this port 80, 443, etc.
# You may edit this if you like, but note:
# Connections are unencrypted. You should only listen on a secure network
# interface.
# The server can only accept a single socket. You cannot specify more than
# one ListenStream address.
ListenStream=[::1]:5000
Accept=no
FreeBind=yes

Save the file and close your editor.

Enable the RT socket

Start the server socket and enable it on future boots:

sudo systemctl enable --now rt-server.socket

Set up RT’s web server

Fully documenting how to set up a web server configured for your network with encryption is outside the scope of this guide. This page only highlights the configuration you need to pass web requests onto the RT FastCGI service.

nginx

Make sure you have the necessary packages installed:

Debian/Ubuntu Red Hat/Fedora/CentOS
sudo apt install nginx
sudo dnf install nginx

In the http server configuration blocks where you want to serve RT, add a configuration block like this, above any other location blocks:

# The location path should match the WebPath in your RT site configuration.
location / {
  include /etc/nginx/fastcgi.conf;
  # SCRIPT_NAME should match RT's WebPath, without a trailing slash.
  # This means when WebPath is /, it's the empty string "".
  fastcgi_param SCRIPT_NAME "";
  # This network location should match the ListenStream in rt-server.socket.
  fastcgi_pass localhost:5000;
}

If you're doing a fresh install, the default http server configuration location varies by distribution:

Debian/Ubuntu Red Hat/Fedora/CentOS
Edit /etc/nginx/sites-available/default Add a new file /etc/nginx/default.d/rt-server.conf

After you’ve edited your configuration, load it by running:

sudo systemctl reload nginx

Once this is done you can skip ahead to Verify the web interface.

Apache httpd

Make sure you have the necessary packages installed, and the proxy_fcgi module enabled:

Debian/Ubuntu Red Hat/Fedora/CentOS
sudo apt install apache2
sudo a2enmod proxy
sudo a2enmod proxy_fcgi
sudo dnf install httpd

In the <VirtualHost> configuration blocks where you want to serve RT, add a configuration block like this, above any other Alias, ProxyPass, or ScriptAlias lines:

# SCRIPT_NAME should match RT's WebPath, without a trailing slash.
# This means when WebPath is /, it's the empty string "".
ProxyFCGISetEnvIf true SCRIPT_NAME ""
# The location path in the first argument should match the WebPath in your
# RT site configuration.  The network location after fcgi:// should match
# the ListenStream in rt-server.socket.
ProxyPass / fcgi://localhost:5000/

If you're doing a fresh install, the default <VirtualHost> configuration location varies by distribution and whether or not you're using HTTPS:

Debian/Ubuntu Red Hat/Fedora/CentOS
With HTTPS: /etc/apache2/sites-available/default-ssl.conf

With plain HTTP: /etc/apache2/sites-available/000-default.conf

With HTTPS: /etc/httpd/conf.d/ssl.conf

With plain HTTP: Write a new file /etc/httpd/conf.d/vhost_RT.conf

After you’ve edited your configuration, load it by running:

Debian/Ubuntu Red Hat/Fedora/CentOS
sudo systemctl reload apache2
sudo systemctl reload httpd

Verify the web interface

You should be able to visit your web server in your browser, and be presented with RT’s login screen. You should be able to log in with username root and the password you set previously.

If the web server returns a 502 Bad Gateway response, it's having trouble connecting to the RT FCGI server. Check the error logs for your web server:

nginx Apache on Debian/Ubuntu httpd on Red Hat/Fedora/CentOS
/var/log/nginx/error.log
/var/log/apache2/error.log
/var/log/httpd/error_log

For all kinds of errors, you can check the logs for the RT FCGI server and its socket:

sudo journalctl --unit rt-server.\*

Set up RT’s mail server

RT both can both send and receive ticket updates via email. Unfortunately, there are too many variables to document a useful setup process here: getting this working usually requires creating DNS records, and coordinating with existing mail servers, which will be the main constraint on your setup. Instead this guide provides a brief overview of how the integration works, and where the connection points are that you likely need to work on.

Sending Mail

RT only knows how to send mail by passing it off to another program on the system. It cannot connect or authenticate directly to external mail servers. In the default configuration, RT runs the standard sendmail command. There are configuration options to send mail through different commands if you need.

The most common setup is to install and configure a proper Mail Transfer Agent (MTA) like Postfix or Exim, and then configure it to send mail to the wider Internet as you need. This works well because the MTAs are robust and well-tested; they have flexible configuration to let you send mail out by relaying to other mail servers you specify with optional authentication; and most distributions install one by default anyway. The only hard part is configuring the MTA to send mail following your site’s policies.

Other software is available that provides a slimmer version of the sendmail command that connects to an external mail server for you, like msmtp. These programs are usually easier to configure than an MTA, but they often lose email permanently if they can’t connect to the external server at the time it’s sent. (MTAs keep email queued locally until they successfully deliver it to another server.)

Receiving email

RT installs a command called rt-mailgate that receives an email on standard input and posts it to RT’s REST web interface, where it gets saved in the database and added to a ticket. You need to arrange for a way to send incoming email to this command.

The most common setup is to have an MTA on the same box as RT receive email directly, and then set up mail aliases that call this command when mail comes in. Example /etc/aliases entries look like:

rt: "|/opt/rt5/bin/rt-mailgate --queue general --action correspond --url https://rt.yourdomain.example.com/"
rt-comment: "|/opt/rt5/bin/rt-mailgate --queue general --action comment --url https://rt.yourdomain.example.com/"

This works well because, again, you’re probably running an MTA anyway; and the MTA can hold and queue mail if it comes in while RT is down for any reason, giving you a buffer against downtime.

Another common option is to periodically run a tool that fetches mail using a protocol like IMAP, like fetchmail or getmail, and passes it on to rt-mailgate. This is less common because it requires setting up another tool to run, and securely storing another set of mail server credentials. But it is useful when local policy prevents the RT server from receiving email directly.

This is much less common, but it might help to know that rt-mailgate doesn’t have to run on the same system as RT itself. It just needs to be able to connect to RT’s web interface. If you don’t have any other options, you can install the RT software on a different system that receives email, and configure that system to run rt-mailgate and pass it on to the RT server. To do that, install the RT::Client::CLI Perl package on the mail server. For example, you could install cpanminus, then run:

cpanm --sudo RT::Client::CLI

Set up RT’s background jobs

Create a file /etc/cron.d/rt with the following content. You may edit all of the time fields as you see fit. Refer to the crontab(5) man page for details about their definitions.

# Update the fulltext index with new ticket data
*/3 *   *   *   *   rt    /opt/rt5/sbin/rt-fulltext-indexer
# Email out dashboards that users have subscribed to
1   *   *   *   *   rt    /opt/rt5/sbin/rt-email-dashboards
# Clean old sessions from the database
50  3   *   *   *   rt    /opt/rt5/sbin/rt-clean-sessions --older 8d
# Email out weekly digests for users who have requested it
50  4   *   *   Mon rt    /opt/rt5/sbin/rt-email-digest -m weekly
# Email out daily digests for users who have requested it
50  5   *   *   *   rt    /opt/rt5/sbin/rt-email-digest -m daily

Set up RT

If you’ve gotten this far, congratulations, your RT install is really done now. You can start setting up RT with users, groups, queues, and business logic. Head back to the main page to start exploring those topics.