Installing OpenSRF
==================

Preamble: referenced user accounts
----------------------------------

In subsequent sections, we will refer to a number of different accounts, as
follows:

  * Linux user accounts:
    ** The *user* Linux account is the account that you use to log onto the
       Linux system as a regular user.
    ** The *root* Linux account is an account that has system administrator
       privileges. On Debian you can switch to this account from
       your *user* account by issuing the `su -` command and entering the
       password for the *root* account when prompted. On Ubuntu you can switch
       to this account from your *user* account using the `sudo su -` command
       and entering the password for your *user* account when prompted.
    ** The *opensrf* Linux account is an account that you will create as part
       of installing OpenSRF. You can switch to this account from the *root*
       account by issuing the `su - opensrf` command.

Download and unpack the code
----------------------------

Issue the following commands as the *user* Linux account.

1. Acquire a stable release tarball from https://evergreen-ils.org/opensrf-downloads/
+
[source, bash]
------------------------------------------------------------------------------
wget https://evergreen-ils.org/downloads/opensrf-OSRFVERSION.tar.gz
------------------------------------------------------------------------------
+
[NOTE]
Developers can find the full source code at the OpenSRF Git repository:
http://git.evergreen-ils.org/?p=OpenSRF.git
+
2. Unpack the tarball, and move into that directory:
+
[source, bash]
------------------------------------------------------------------------------
tar -xvf opensrf-OSRFVERSION.tar.gz
cd opensrf-OSRFVERSION/
------------------------------------------------------------------------------

Installing prerequisites
------------------------

OpenSRF has a number of prerequisite packages that must be installed
before you can successfully configure, compile, and install OpenSRF.
On Debian and Ubuntu, the easiest way to install these prerequisites
is to use the Makefile.install prerequisite installer.

Issue the following commands as the *root* Linux account to install 
prerequisites using the Makefile.install prerequisite installer, substituting 
your operating system identifier for <osname> below:

[source, bash]
---------------------------------------------------------------------------
apt-get install make
make -f src/extras/Makefile.install <osname>
---------------------------------------------------------------------------

Well-tested values for <osname> include:

  * `debian-bullseye` for Debian 11
  * `debian-buster` for Debian 10
  * `ubuntu-focal` for Ubuntu 20.04
  * `ubuntu-jammy` for Ubuntu 22.04

Patches and suggestions for improvement from users of these distributions,
or others, are welcome!

When the prerequisite installer reaches the Perl module stage, you may 
be prompted for configuration of Comprehensive Perl Archive Network (CPAN)
on your server. You can generally accept the defaults by pressing <return>
for all of the prompts, except for the country configuration.

Preamble: Developer instructions
--------------------------------

[NOTE]
Skip this section if you are using an official release tarball downloaded
from https://evergreen-ils.org/opensrf-downloads/

Developers working directly with the source code from the Git repository,
rather than an official release tarball, must install some extra packages
and perform one step before they can proceed with the `./configure` step.

As the *root* Linux account, install the following packages:

  * autoconf
  * automake
  * libtool

As the *user* Linux account, issue the following command in the OpenSRF
source directory to generate the configure script and Makefiles:

[source, bash]
------------------------------------------------------------------------------
autoreconf -i
------------------------------------------------------------------------------

Configuration and compilation instructions
------------------------------------------

Use the `configure` command to configure OpenSRF, and the `make` command to
build OpenSRF. The default installation prefix (PREFIX) for OpenSRF is
`/opensrf/`.

If you are building OpenSRF for Evergreen, issue the following commands as
the *user* Linux account to configure and build OpenSRF:

[source, bash]
---------------------------------------------------------------------------
./configure --prefix=/openils --sysconfdir=/openils/conf
make
---------------------------------------------------------------------------

By default, OpenSRF includes C, Perl, and JavaScript support.

If you are planning on proxying WebSockets traffic (see below), you
can add `--with-websockets-port=443` to specify that WebSockets traffic
will be going through port 443. Without that option, the default port
is 7682.

Installation instructions
-------------------------

1. Once you have configured and compiled OpenSRF, issue the following
   command as the *root* Linux account to install OpenSRF:
+
[source, bash]
---------------------------------------------------------------------------
make install
---------------------------------------------------------------------------

Create and set up the opensrf Unix user environment
---------------------------------------------------

This user is used to start and stop all OpenSRF processes, and must own all
files contained in the PREFIX directory hierarchy. Issue the following
commands as the *root* Linux account to create the `opensrf` user and set up
its environment, substituting <PREFIX> with the value you passed to `--prefix`
in your configure command:

.Creating the `opensrf` user
[source, bash]
---------------------------------------------------------------------------
useradd -m -s /bin/bash opensrf
echo "export PATH=\$PATH:/<PREFIX>/bin" >> /home/opensrf/.bashrc
passwd opensrf
chown -R opensrf:opensrf /<PREFIX>
---------------------------------------------------------------------------

Define your public and private OpenSRF domains
----------------------------------------------

For security purposes, OpenSRF uses Jabber domains to separate services
into public and private realms. Throughout these instructions, we will use
the example domains `public.localhost` and `private.localhost`. 

On a single-server system, the easiest way to define public and private
domains is to define separate hostnames by adding entries to the `/etc/hosts`
file. Here are entries that you could add to a stock `/etc/hosts` file for our
example domains:

.Example added entries for `/etc/hosts`
[source, bash]
---------------------------------------------------------------------------
127.0.1.2	public.localhost	public
127.0.1.3	private.localhost	private
---------------------------------------------------------------------------

Adjust the system dynamic library path
--------------------------------------

Add `<PREFIX>/lib/` to the system's dynamic library path, and then run
`ldconfig` as the *root* Linux account.

On Debian and Ubuntu systems, run the following commands as the *root*
Linux account:

.Adjusting the system dynamic library path
[source, bash]
---------------------------------------------------------------------------
echo <PREFIX>/lib > /etc/ld.so.conf.d/opensrf.conf
ldconfig
---------------------------------------------------------------------------

On most other systems, you can add these entries to `/etc/ld.so.conf`, or
create a file within the `/etc/ld.so.conf.d/` directory, and then run
`ldconfig` as the *root* Linux account.

Configure the ejabberd server
-----------------------------

[TIP]
=====
It is recommended to disable the apparmor profile for ejabberd on *Ubuntu* before
continuing.  If you are installing on any version of *Ubuntu*, run the following
commands as the *root* Linux account:

[source,bash]
---------------------------------------------------------------------------
ln -s /etc/apparmor.d/usr.sbin.ejabberdctl /etc/apparmor.d/disable/
apparmor_parser -R /etc/apparmor.d/usr.sbin.ejabberdctl
---------------------------------------------------------------------------
=====

OpenSRF requires an XMPP (Jabber) server. For performance reasons, ejabberd is
the Jabber server of choice for the OpenSRF project. In most cases, you only
have to make a few changes to the default configuration file to make ejabberd
work for OpenSRF. 

1. Stop ejabberd before making any changes to its configuration by issuing the
   following command as the *root* Linux account:
+
.Stopping ejabberd
[source, bash]
---------------------------------------------------------------------------
systemctl stop ejabberd.service
---------------------------------------------------------------------------
+
2. Edit the ejabberd config file.
+
(Debian Buster / Ubuntu Focal) Ejabberd 18.x::
Open `/etc/ejabberd/ejabberd.yml` and make the following
changes:
  a. Define your public and private domains in the `hosts` directive. For
   example:
+
[source, bash]
---------------------------------------------------------------------------
hosts:
  - "localhost"
  - "private.localhost"
  - "public.localhost"
---------------------------------------------------------------------------
+
  b. Change `starttls_required` to false
  c. Change `auth_password_format` to plain
  d. Change `shaper:` `normal` and `fast` values to 500000
  e. Increase the `max_user_sessions:` `all:` value to 10000
  f. Comment out the `mod_offline` directive
+
-----------------------
##mod_offline:
    ##access_max_user_messages: max_user_offline_messages
-----------------------
+
  g. Uncomment or add the `mod_legacy_auth` directive under the `modules:` section
+
-----------------------
mod_legacy_auth: {}
-----------------------
+
(Debian Bullseye / Ubuntu Jammy) Ejabberd 21.x::
Open `/etc/ejabberd/ejabberd.yml` and make the following
changes:
  a. Define your public and private domains in the `hosts` directive. For
   example:
+
[source, bash]
---------------------------------------------------------------------------
hosts:
  - localhost
  - private.localhost
  - public.localhost
---------------------------------------------------------------------------
+
  b. Change `starttls_required` to false
  c. Change `auth_password_format` to plain
  d. Change all `shaper:` `normal` and `fast` values to 500000
  e. Increase the `shaper_rules:` `max_user_sessions:` value to 10000
  f. Comment out the `shaper_rules:` `max_user_offline_messages:` values
+
-----------------------
##max_user_offline_messages:
  ##5000: admin
  ##100: all
-----------------------
+
  g. Comment out the `mod_offline` directive
+
-----------------------
##mod_offline:
    ##access_max_user_messages: max_user_offline_messages
-----------------------
+
  h. Add the `mod_legacy_auth` directive under the `modules:` section
+
-----------------------
mod_legacy_auth: {}
-----------------------
+
3. Restart the ejabberd server to make the changes take effect:
+
.Starting ejabberd
[source, bash]
---------------------------------------------------------------------------
systemctl start ejabberd.service
---------------------------------------------------------------------------

Create the OpenSRF Jabber users
-------------------------------

On each domain, you need two Jabber users to manage the OpenSRF communications:

  * a `router` user, to whom all requests to connect to an OpenSRF service
    will be routed; this Jabber user must be named `router`
  * an `opensrf` user, which clients use to connect to OpenSRF services; this
    user can be named anything you like

Create the Jabber users by issuing the following commands as the *root* Linux
account. Substitute `<password>` for your chosen passwords for each user
respectively:

.Creating the OpenSRF Jabber users
[source, bash]
---------------------------------------------------------------------------
ejabberdctl register router private.localhost <password>
ejabberdctl register opensrf private.localhost <password>
ejabberdctl register router public.localhost <password>
ejabberdctl register opensrf public.localhost <password>
---------------------------------------------------------------------------

Update the OpenSRF configuration files
--------------------------------------

About the OpenSRF configuration files
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
There are several configuration files that you must update to make OpenSRF
work. SYSCONFDIR is `/opensrf/etc` by default, or the value that you passed to
`--sysconfdir` during the configuration phase.

  * `SYSCONFDIR/opensrf.xml` - this file lists the services that this
    OpenSRF installation supports; if you create a new OpenSRF service,
    you need to add it to this file.
      ** The `<hosts>` element at the bottom of the file lists the services
         that should be started for each hostname. You can force the system
         to use `localhost`, so in most cases you will leave this section
         as-is.
    
  * `SYSCONFDIR/opensrf_core.xml` - this file lists the Jabber connection
    information that will be used for the system, as well as determining
    logging verbosity and defining which services will be exposed on the
    HTTP gateway.

  * `~/.srfsh.xml` - this file gives a Linux account the ability to use
    the `srfsh` interpreter to communicate with OpenSRF services.

Updating the OpenSRF configuration files
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  1. As the *opensrf* Linux account, copy the example configuration files
     to create your locally customizable OpenSRF configuration files:
+
.Copying the example OpenSRF configuration files
[source, bash]
---------------------------------------------------------------------------
cd SYSCONFDIR
cp opensrf_core.xml.example opensrf_core.xml
cp opensrf.xml.example opensrf.xml
---------------------------------------------------------------------------
+
  2. Edit the `SYSCONFDIR/opensrf_core.xml` file to update the four username
     / password pairs to match the Jabber user accounts you just created:

    a. `<config><opensrf>` = use the private Jabber `opensrf` user
    b. `<config><gateway>` = use the public Jabber `opensrf` user
    c. `<config><routers><router>` = use the public Jabber `router` user
    d. `<config><routers><router>` = use the private Jabber `router` user
  3. Create a `.srfsh.xml` file in the home directory of each user
     that you want to use `srfsh` to communicate with OpenSRF services. For
     example, to enable the *opensrf* Linux account to use `srfsh`:
    a. `cp SYSCONFDIR/srfsh.xml.example ~/.srfsh.xml`
    b. Open `~/.srfsh.xml` in your text editor of choice and update the
       password to match the password you set for the Jabber `opensrf` user
       at the `private.localhost` domain.

Starting and stopping OpenSRF services
--------------------------------------

To start all OpenSRF services with a hostname of `localhost`, issue the
following command as the *opensrf* Linux account:

[source, bash]
---------------------------------------------------------------------------
osrf_control --localhost --start-all
---------------------------------------------------------------------------

To stop all OpenSRF services with a hostname of `localhost`, issue the
following command as the *opensrf* Linux account:

[source, bash]
---------------------------------------------------------------------------
osrf_control --localhost --stop-all
---------------------------------------------------------------------------

Testing the default OpenSRF services
------------------------------------

By default, OpenSRF ships with an `opensrf.math` service that performs basic
calculations involving two integers. Once you have started the OpenSRF
services, test the services as follows:

1. Start the `srfsh` interactive OpenSRF shell by issuing the following
   command as the *opensrf* Linux account:
+
.Starting the `srfsh` interactive OpenSRF shell
[source, bash]
---------------------------------------------------------------------------
srfsh
---------------------------------------------------------------------------
+
2. Issue the following request to test the `opensrf.math` service:
+
[source, bash]
---------------------------------------------------------------------------
srfsh# request opensrf.math add 2,2
---------------------------------------------------------------------------
+
You should receive the value `4`.

Websockets installation instructions
------------------------------------

1. Install websocketd (latest stable release from http://websocketd.com/)
+
.(Debian, Ubuntu)
[source,bash]
---------------------------------------------------------------------------
cd /tmp
wget 'https://github.com/joewalnes/websocketd/releases/download/v0.3.0/websocketd-0.3.0-linux_amd64.zip'
unzip websocketd-0.3.0-linux_amd64.zip
sudo cp websocketd /usr/local/bin/
---------------------------------------------------------------------------
+
2. Run websocketd
+
Choose option a or b, below.
+
[NOTE]
===========================================================================
websocketd does not offer a configurable inactivity timeout, meaning
websocket client connections will persist until each client disconnects
or the service is restarted.  However, a timeout can be achieved with
the use of a proxy (option 'a' below).  A proxy also allows websocketd
to be exposed to web clients on port 443 instead of its internal port, 
which may simplify firewall configuration.
===========================================================================
+
a. Run websocketd as 'opensrf'
+
[NOTE]
===========================================================================
This choice requires one of the proxy configurations mentioned below.
===========================================================================
+
.(Debian, Ubuntu)
[source,bash]
---------------------------------------------------------------------------
/usr/local/bin/websocketd --port 7682 /openils/bin/osrf-websocket-stdio &

# Other useful command line parameters include:
# --loglevel debug|trace|access|info|error|fatal
# --maxforks <n>
# --sameorigin=true
# --origin=host[:port][,host[:port]...]

# See https://github.com/joewalnes/websocketd/blob/master/help.go
---------------------------------------------------------------------------
+
b. Run websocketd without a proxy
+
.(Debian, Ubuntu)
[source,bash]
---------------------------------------------------------------------------
sudo -b /usr/local/bin/websocketd --port 7682 --ssl --sslcert=/etc/apache2/ssl/server.crt \
     --sslkey=/etc/apache2/ssl/server.key /openils/bin/osrf-websocket-stdio
---------------------------------------------------------------------------

Optional Systemd Setup
~~~~~~~~~~~~~~~~~~~~~~

Websocketd is a standalone program with no daemon mode, but can be implemented as a systemd service.

Copy <PREFIX>/examples/websocket-osrf.service.example into file /lib/systemd/system/websocketd-osrf.service

Then add & start the service.

[source,bash]
--------------------------------------
sudo systemctl daemon-reload
sudo systemctl enable websocketd-osrf
sudo systemctl start websocketd-osrf
--------------------------------------

Optional: Using a web proxy (Apache 2.4 and above)
--------------------------------------------------
When the OpenSRF HTTP Translator runs behind a proxy, Apache must be 
configured to read the IP address of the originating client instead
of the proxy IP address.  

1. Enable mod_remoteip
+
[source,bash]
---------------------------------------------------------------------------
sudo a2enmod remoteip
---------------------------------------------------------------------------
+
2. Enable remote IP settings by uncommenting and modifying as needed the 
   Apache configuration variables starting with RemoteIP* in the sample Apache
   configuration file opensrf.conf.
+
3. Configure Apache to listen on port 7080 for HTTP and port 7443 for HTTPS
   and ensure that it is not listening on ports 80 and 443, then restart Apache.
+
4. If you didn't run `configure` with the `--with-websockets-port=443` option,
   edit `<PREFIX>/javascript/opensrf_ws.js` and `<PREFIX>/javascript/opensrf_ws_shared.js`
   and change
+
[source, javascript]
---------------------------------------------------------------------------
var WEBSOCKET_PORT_SSL = 7682;
---------------------------------------------------------------------------
+
to
+
[source, javascript]
---------------------------------------------------------------------------
var WEBSOCKET_PORT_SSL = 443;
---------------------------------------------------------------------------


Optional: Using NGINX as a proxy
--------------------------------
NGINX can be used to proxy HTTP, HTTPS, and WebSockets traffic. Among other
reasons, this can be useful for Evergreen setups that want to have both
HTTPS and secure WebSockets traffic both go through port 443 while using
two Apache instances (one for the WebSockets gateway and one for the more
memory-intensive TPAC pages).

The following instructions are a guide for setting this up on Debian
and Ubuntu systems, but expect general familiarity with various system
administration and network tasks.  The steps should be run as the *root*
Linux account, and assume that you already followed the instructions
for installing WebSockets support.

1. Install NGINX if not already present:
+
[source, bash]
---------------------------------------------------------------------------
apt-get install nginx
---------------------------------------------------------------------------
+
2. Copy the example NGINX configuration file into place and remove default.
+
[source, bash]
---------------------------------------------------------------------------
cd /path/to/opensrf-OSRFVERSION
cp examples/nginx/osrf-ws-http-proxy /etc/nginx/sites-available/
ln -s /etc/nginx/sites-available/osrf-ws-http-proxy /etc/nginx/sites-enabled/osrf-ws-http-proxy
rm /etc/nginx/sites-enabled/default
---------------------------------------------------------------------------
+
3. Edit `/etc/nginx/sites-available/osrf-ws-http-proxy` to set the location
   of the SSL certificate and private key.
4. Generate a dhparam file in the directory specified in the nginx config.
+
[source, bash]
---------------------------------------------------------------------------
# Default config stores dhparam.pem in the Apache2 ssl directory.
openssl dhparam -out /etc/apache2/ssl/dhparam.pem 2048
---------------------------------------------------------------------------
+
5. Start NGINX
+
[source, bash]
---------------------------------------------------------------------------
/etc/init.d/nginx start
---------------------------------------------------------------------------

Optional: Using HAProxy as a proxy
----------------------------------
HAProxy can also be used to proxy HTTP, HTTPS, and WebSockets traffic
as an alternative to NGINX.

The following instructions are a guide for setting this up on Debian
and Ubuntu systems, but expect general familiarity with various system
administration and network tasks.  The steps should be run as the *root*
Linux account, and assume that you already followed the instructions
for installing WebSockets support.

1. Install HAProxy if not already present:
+
[source, bash]
---------------------------------------------------------------------------
apt-get install haproxy
---------------------------------------------------------------------------
+
2. Append the example HAProxy to `haproxy.cfg`.
+
[source, bash]
---------------------------------------------------------------------------
cd /path/to/opensrf-OSRFVERSION
cat examples/haproxy/osrf-ws-http-proxy >> /etc/haproxy/haproxy.cfg
---------------------------------------------------------------------------
+
3. Edit `/etc/haproxy/haproxy.cfg` to set the location
   of the PEM file containing the SSL certificate and private key.
4. Start HAProxy.
+
[source, bash]
---------------------------------------------------------------------------
/etc/init.d/haproxy start
---------------------------------------------------------------------------

Getting help
------------

Need help installing or using OpenSRF? Join the mailing lists at
http://evergreen-ils.org/communicate/mailing-lists/ or contact us 
on the Freenode IRC network on the #evergreen channel.