Configuration Management on FreeBSD with CFEngine - Part I

Administrating a couple hundreds of systems might easily become a tedious tasks, even for the very experienced system administrators.

Keeping a mindset where you automate every possible aspect of an Operating System and the tasks you do on it, is a must when you are responsible for that many (or even more) systems.

The FreeBSD Ports Collection is a great thing to use when installing software distributions, but installing multiple ports on several hundreds of machines is not a pleasant tasks - it is time consuming, needs user attention, and not to forget - things might blow up at some point.

Having central systems that take care of building the packages and then distributing them to all systems in the domain, not only saves the time needed to build a package, but also allows us to be more flexible in the package deployment, in the sense of having different environments - a test environment where we test the newly built packages, which we can later safely deploy to our production environment.

The following handbook is organized into different chapters, each of them outlining the steps needed to setup a certain part of our environment.

The first chapter takes a look into Tinderbox - a system for testing and building packages under FreeBSD.

We will be using Tinderbox for building our packages, which will then be distributed to all the clients.

The second chapter is about CFEngine 3, which describes how to install and configure Cfengine 3 and put it's configuration under revision control with Git.

We will see how to automate a configuration deployment to our systems with CFEngine 3, and also do package installations and upgrades.

In the last chapter we will take a look into different examples of using CFEngine 3, and how to accomplish specific tasks with it.


Installation of CFEngine

In this chapter we will take a look into CFEngine 3 - a configuration management framework for configuring and maintaining UNIX and UNIX-like systems.

Anything you can do on a system in general can be automated via CFEngine 3, which makes it perfect for deploying mass changes to a lot of systems.

CFEngine 3 is a suite of programs, each of them responsible for executing a specific task.

  • cf-agent - Active agent
  • cf-execd - CFEngine Executor / Scheduler
  • cf-know - Knowledge modeling agent
  • cf-monitord - Passive monitoring agent
  • cf-promises - Promise validator
  • cf-runagent - Remote run agent
  • cf-serverd - CFEngine Server
  • cf-report - Self-knowledge extractor
  • cf-key - Key generation tool
  • cf-hub - Data aggregator

CFEngine 3 uses the client/server model, in which one (or more) systems act as the master servers (also referred as the policy servers), which keeps the promises and files which will be in turn distributed to the clients (also referred as the agents).

Using CFEngine 3 you can keep your systems in a desired state. When an agent is running it will try to conform to the promises it made, in order to keep the system to the state we want it to be.

Since the following handbook is just a basic introduction to CFEngine 3 and it's capabilities, you are also advised to have a look at the online documentation of CFEngine 3 for more in-depth explanation of Cfengine 3 and it's features.

Now let's go ahead and prepare our systems for CFEngine 3.

Building the CFEngine 3 package

Now that we have a Tinderbox system installed and configured already for us, for all of the package builds from now on, we are going to make use of Tinderbox.

So, now go ahead and build your sysutils/cfengine3 package in Tinderbox, as it was described in the previous chapter.

After building the package, we will use it in order to install it on the CFEngine 3 master server, and it's clients.

The version of the CFEngine 3 package that we use in this handbook is cfengine-3.1.4_1

We are also going to put the CFEngine 3 configuration under version control with Git, so we will need the FreeBSD package for Git as well.

So build the devel/git package in your Tinderbox as well, so we can later install it.

The version of Git we use in this handbook is git-

Preparing the CFEngine 3 master servers

In this step we will see how to install and prepare the CFEngine 3 master server, which is also referred to as the policy server.

This is the server which keeps the promises and files, which will be in turn distributed to the clients, also referred to as the agents.

The CFEngine 3 master server will be installed on a FreeBSD jail. So, the first step is to create a new FreeBSD jail, and then login to it.

I would also advise you on using sysutils/ezjail for managing your FreeBSD jails, since it provides easy to use interface to your FreeBSD jails.

Also, we will make use of the ezjail flavours later on so that newly created jails will also be configured with CFEngine 3, once they are installed.

For more information on sysutils/ezjail, please refer to the official web page of ezjail.

Using the ezjail flavours we can have different types of configurations for different jails.

In the following handbook we will have at least two ezjail flavours - one for the test environment and one for our production environment.

Each of these ezjail flavours will in turn install the CFEngine 3 package we have built already, and the corresponding CFEngine 3 configuration for that environment.

Now, create the jails, which will be our CFEngine 3 policy servers. In this handbook we will have two CFEngine 3 master servers - one for the test environment, and one for our production environment.

$ sudo ezjail-admin create cfengine-test
$ sudo ezjail-admin create cfengine

For more information on the syntax of ezjail-admin(1), please refer to it's manual page.

Installing the CFEngine 3 package

Now, that we have a FreeBSD jail, let's install the CFEngine 3 package that we have built already in Tinderbox.

$ sudo ezjail-admin console cfengine

In the Tinderbox chapter we mentioned something about transferring the Tinderbox packages to a local FTP server using either a Tinderbox hook, or a simple cron job that rsyncs our packages to the FTP server.

In the next command we are actually using the local FTP package repository we have in place already.

So now let's install the Git and CFEngine 3 packages on our CFEngine 3 master server.

# pkg_add -r git-
# pkg_add -r cfengine-3.1.4_1

The CFEngine 3 package installs in /usr/local, but CFEngine 3 needs to have a trusted work directory which is set to /var/cfengine - this is also the place where the CFEngine promises and files are stored.

But before creating the CFEngine work directory in /var/cfengine, we are going to put all the CFEngine configuration under version control with Git.

Preparing CFEngine Git repository

Assuming you have read the handbook for installing and configuring Git on your FreeBSD system, we will now create the Git repository for CFEngine on our Git server.

On the Git server:

$ cd /home/git/public && mkdir cfengine
$ cd cfengine && git init --bare --shared

Now we can go to the next chapter, where we'll start preparing the basic CFEngine 3 configuration.

Written on April 16, 2011