Installation of Kofa
####################

These are generic installation instructions for the WAeUP_ ``Kofa``
student information portal and customized versions thereof.

For specific aspects of this package please refer to the local README
file.

Please note, that **only Linux-based installs** are described and actively
supported. We recommend use of Debian_ / Ubuntu_.

.. note:: This means we do not actively support Windows(tm)!

.. contents::


Prerequisites
*************

The Kofa packages are based on `Grok`_, which is a Python_
framework for agile webapplication development. Grok_ itself is based
on `Zope`_.

Both, Grok_ and Zope_, are written in the `Python`_ programming
language (with parts written in C). You therefore have to have
`Python`_ installed (including `Python`_ headers).


Preparing the System
====================

For a `Kofa` install we need of course `Python`_ installed as well as
some standard developer tools like C compilers, C libraries, etc.

What you need (Debian/Ubuntu package names in brackets):

* Python 2.7 (``python2.7``)

* Python 2.7 development files (``python2.7-dev``)

* A C-Compiler with basic C developer libraries (``build-essential``)

* A subversion client (``svn``)

* enscript (``enscript``) [optional]

  This is only needed if you want test coverage reports.

All these packages can be installed on Debian systems like this::

  $ sudo apt-get install python2.7 python2.7-dev svn \
                         build-essential enscript

Afterwards you should be able to enter::

  $ python2.7

at the commandline and get a Python_ prompt. Quit the interpreter
pressing <CTRL-D>.


Installing `virtualenv`
=======================

We strongly suggest use of `virtualenv`_ to create Python_ sandboxes
where you can run your code without touching any other installations.

`virtualenv`_ is also the only possibility to build a `Kofa` install
without the need of superuser permissions. In short: `virtualenv`_
will make your life easier. Use it.

Detailed install instructions for `virtualenv`_ can be found on
http://www.virtualenv.org/en/latest/virtualenv.html#installation.

The short way for a user install (no superuser perms required) is like
this::

  $ curl -O https://pypi.python.org/packages/source/v/virtualenv/virtualenv-1.11.4.tar.gz#md5=9accc2d3f0ec1da479ce2c3d1fdff06e
  $ tar xvfz virtualenv-1.11.4.tar.gz
  $ cd virtualenv-1.11.4
  $ python virtualenv.py py27

Superusers can install `virtualenv`_ system-wide much easier. On
`Debian`_/`Ubuntu`_ systems where you have superuser permissions, you
can install `virtualenv`_ via::

  $ sudo apt-get install python-virtualenv

*Or*, if `pip` is installed already (superusers can install it via ``sudo apt-get install python-pip`` on `Debian`_/`Ubuntu`_)::

  $ sudo pip install virtualenv

In the following, when referring to the `virtualenv` command, we talk
of the system-wide available command or the ``virtualenv.py`` script
installed for a local user. In the latter case you should read::

  $ virtualenv ...

as::

  $ python /path/to/my/virtualenv.py ...


Creating a sandbox
==================

After installing `virtualenv`_ you can install local sandboxes like
this::

  $ virtualenv --no-site-packages py27

where ``py27`` is a directory in the filesystem where your sandbox
will be created. `virtualenv` will also create this directory for
you. It is recommended to use a sandbox name that is short and at the
same time reflects the Python version used (here: Python 2.7).

By passing the ``no-site-packages`` switch we tell `virtualenv` to
provide us a clean environment without any extra-packages installed
systemwide. More recent versions of `virtualenv`_ have this option set
by default.

Another often used option of `virtualenv`_ is ``-p``. With ``-p`` you
can tell `virtualenv`_ which Python executable to use as base for the
new sandbox.

You now can activate the sandbox by doing::

  $ source py27/bin/activate
  (py27)$

You will notice that the input prompt changes.

To deactivate the sandbox at any time, enter::

  (py27)$ deactivate

and the prompt will be the same as before the activation.

For the following steps make sure the sandbox is active.


Creating a working place
========================

In the sandbox we now create our real working
environment. To do this, we change to the sandbox and checkout the
sources of Kofa from the subversion server::

  (py27)$ cd mysandbox/
  (py27)$ svn co https://svn.waeup.org/repos/main/waeup.kofa/trunk kofa-trunk

where ``kofa-trunk`` is only a name we've chosen here to make clear
where the sources come from. In this case we are installing the Kofa base
package.

The command should fetch the Kofa base package sources for you and
put it in the directory ``kofa-trunk/``.

Now enter the new directory::

  (py27)$ cd kofa-trunk


You can run ``Kofa`` with a single client (useful for evaluation,
development, etc.) or with multiple clients running in parallel. The
latter setup is useful in productive environments with many thousands
users. We will cover both setup types one after another.


Part 1: Single Client Setup
***************************

Preparing the build
===================

In the sources directory (``kofa-trunk/``) you have to prepare the
project to fetch needed components (eggs), compile C-code parts,
etc. This steip will not touch any external projects::

  (py27)$ python bootstrap.py

This will generate some directories and the ``buildout`` script in
``bin/`` for us. This step must be executed only once for each
instance.

You can now deactivate the sandbox::

  (py27)$ deactivate

Now we can do the real build by triggering::

  $ bin/buildout

If this is your first install of some Grok-related project, this step
will need some time as lots of sources have to be fetched, many
components must be compiled, etc.

This step must be redone whenever you change something in
``buildout.cfg`` or ``setup.py``.

Note that if you have more than one sandbox for a Zope-based web
application, it will probably make sense to share the eggs between the
different sandboxes.  You can tell ``zc.buildout`` to use a central
eggs directory by creating ``~/.buildout/default.cfg`` with the
following contents::

  [buildout]
  eggs-directory = /home/bruno/buildout-eggs


Start the instance
==================

You should be able now to start the created instance by doing::

  $ bin/zopectl fg

Alternatively you can do::

  $ bin/paster serve parts/etc/deploy.ini

The port numbers where Kofa is running on your server are defined in
buildout.cfg under [kofa_params].

If you now point a browser to the right port on your server, for example ::

  localhost:8080

you should get a login pop-up, where you can login as superuser with
``grok`` and ``grok`` as username/password (Kofa base package only!).

You can stop the instance by pressing <CTRL-C>.

If you are connected and logged in,
you should be able to add the grok-based applications
(such as ``University``) from the menu.

Add an instance of ``University`` and click on the link next to the
then visible entry in the list of installed applications.

Running the tests
=================

The tests are easily run by executing the test runner that's
installed in the ``bin`` directory::

  $ bin/test


Part 2: Deploying Kofa as ZEO install
*************************************

Each ZEO install consists of at least one ZEO server and normally two
or more ZEO clients. While the ZEO server is meant to manage the ZODB
database for clients, the clients connect to the outside world, listen
for request and do the real dataprocessing.

We prepared a `buildout` configuration that sets up one server
configuration and two client configs. This configuration is in
``buildout-zeo.cfg``.

Generating the ZEO setup
========================

To install Kofa ZEO-based you can run `buildout` with the given
(or your own) configuration file like this::

  $ ./bin/buildout -c buildout-zeo.cfg

This should generate all scripts necessary to run servers, clients,
etc.

Starting ZEO servers and clients
================================

First start the server::

  $ ./bin/zeo_server start

Clients can be started by doing::

  $ ./bin/zeo_client1 start
  $ ./bin/zeo_client2 start

This will start both clients in daemon mode.

Instead of ``start`` you can, as usually, start an instance in
foreground (``fg``), etc. You know the drill.


Manually starting ZEO clients
=============================

This is normally not neccessary.

``zeo_clientN`` scripts are basically wrappers around calls to
``bin/paster``. You can bypass this wrapper and start a client
'manually' like this::

  $ ./bin/paster serve --pid-file var/zeo1.pid --daemon \
        pars/etc/zeo1.ini

It is important to give a pid-file as paster otherwise can not start
different clients (they would all refer to the same pid file
`paster.pid` and refuse to start after the first client was started).


Setup (paramters, ports, etc.)
==============================

By default the server will listen on port 8100 for requests from
localhost (not: from the outside world).

You can change ZEO server settings in the ``[zeo_server]`` section
of ``buildout-zeo.conf``. Run `buildout` afterwards.

The clients will listen on port 8081 and 8082. You can change settings
in ``etc/zeo1.ini.in`` and ``etc/zeo2.ini.in`` respectively. Run
buildout after any change.

If you want to change the paster wrapper for any zeo client, you can
edit ``etc/zeo1.conf`` and/or ``etc/zeo2.conf``. Run buildout
afterwards.


Creating new clients
====================

You want more clients to be created by buildout? Easy. Three steps are
neccessary.

1. Create config files in etc/
------------------------------

Each client needs two configuration files:

  - ``etc/zeoN.conf``
       configuring the paster wrapper

  - ``etc/zeoN.ini``
       configuring the runtime config, ports, etc.

Just copy over these files from the already existing zeo1/zeo2 files
and replace ``zeo1`` or ``zeo2`` with your new name.

2. Update buildout-zeo.cfg
--------------------------

Here, inside ``buildourt-zeo.cfg`` also three steps are needed.

* 2.1. Create new .ini and .conf entries

  The .conf and .ini files in etc/ are only templates that have to be
  generated in their really used final location. In buildout-zeo.cfg
  you can care for this by creating a new ``[zeoN_ini]`` and
  ``[zeoN_conf]`` option (replacing ``N`` with a number, of course).

  Just copy over existing entries and replace the mentions of ``zeo1``
  or ``zeo2`` by your ``zeoN``.

* 2.2. Create a new ``zeo_clientN`` entry

  Then you have to create an entry that will generate the
  ``zeo_clientN`` script. Again, just copy over an existing
  ``[zeo_client1]`` entry and replace ``1`` withg your client number.

* 2.3. Register the new sections in ``[buildout]`` section

  When done with the above: add the new section in ``[buildout]``::

    [buildout]
      ...
      <old entries...>
      ...
      zope_conf_zeo_5
      zeo5_ini
      zeo_client5

  depending on how you named your new sections.

3. Rerun ``buildout``
---------------------

When adding or removing client/server instances, make sure to stop all
running servers/clients before rerunning buildout.

To activate the new setup, rerun buildout::

  $ bin/buildout -c buildout-zeo.cfg

This should generate any new clients and remove older ones or just
update configuration files.


Considerations
==============

There are some things in the current buildout-zeo.cfg we might do not
want. It extends the regular ``buildout.cfg`` so that we do not have
to repeat most sections but the ``parts`` in ``[buildout]`` have to be
listed.

We need, however, not everything with a ZEO-deploy that is listed in a
default buildout. We might do not need docs, no profiling, etc. Also a
regular non-ZEO kofactl might not make to much sense. Therefore all
this might be subject to changes.

.. _Debian: http://www.debian.org/
.. _Grok: http://grok.zope.org/
.. _Python: http://www.python.org/
.. _Ubuntu: http://www.ubuntu.com/
.. _virtualenv: http://www.virtualenv.org/en/latest/
.. _WAeUP: https://www.waeup.org/
.. _Zope: http://www.zope.org/
.. _zc.buildout: http://cheeseshop.python.org/pypi/zc.buildout