Define a complete applicative policy

In this section, you will learn:

  • How to define more advanced configuration policies

  • How to link different components, using conditionnal evaluation of methods.

  • How to define and use techniques with parameters

Technique creation

In the first part, we defined a very simple technique that managed a user.

We will go back to the technique editor (Configuration policy → Techniques) and create a technique to configure a small website, that will consist of an nginx webserver installation, a deployment of a minimal configuration and an html file as our website.

Let’s call our new technique "Static website".

Website technique

But this time, we will add parameters to this technique. What are parameters? Just like we saw in the built-in techniques, you can pass parameters to techniques created in the technique editor. They are part of the technique definition, and their value is usable with the ${variable_name} syntax.

We will need two parameters:

  • A path for our site, called webroot

  • A port to listen on, called port

Technique parameters

Then we can start assembling our building blocks. We want:

  • to install nginx and manage the nginx service

  • to add our website content

  • to manage nginx configuration file


We will now map our use case to generic methods in the technique editor.

Nginx installation

Let’s start with the easy part, the nginx setup. We need to install the nginx package.

To do so, we will use a package present method, that will continuously ensure the package is installed. You can skip all parameters except the package name, which is nginx, defaults are ok for us.

Install nginx package

And for the service, use a service started method, with the nginx service name. It will ensure the service is started.

Start nginx service

Website content

File content

Four our website content, we will use a simple File content with the html content of our website. This method can add content to a file, this is the most simple file editing method.

<html><h1>Welcome to Rudder demo!</h1></html>

And put it in our webroot using ${webroot}/index.html.

The enforce parameter specifies if we want to add content (with the false value) or to totally replace the content of the file (with the true value).
HTML index files
This one is only a demo example, a real use case would probably include a separate website deployment process.


We need a second method for our website content, as the default permission for files created by Rudder is 640, owned by root. Our file would hence not be readable by the web server user. To configure specific permissions, use the permissions (non-recursive) method with the following parameters:

  • mode: 644

  • owner: root

  • group: root

File permissions

Nginx configuration file

The last missing part is the web server configuration file. We want to tell nginx to serve our page on the domain given as parameter, and on the given port.

This time we won’t use a simple file content, but a dedicated feature for such cases: templating. It allows providing a base file content, with special markers for parts that will be dynamically computed by the agent at execution. It allows to have node-specific content in a generic file (like the static file server in our example). Rudder’s native templating language is named mustache (you’ll see why very soon).

Templating allows replacing content in a generic template with specific data. The include environement conditions or variables.

It is generally a good practice to enforce the content of file totally, either by templating or file copy than to use file editions, as they define an absolute state and are hence a lot easier to use and more reliable.

Let’s add this configuration file:

server {
    root {{{vars.Static_website.webroot}}};
    listen {{{vars.Static_website.port}}};

In the destination /etc/nginx/conf.d/demo.conf using a file from a mustache string method that allows us providing the template content directly in the web interface.

Templating method

We have now defined all of our configuration states! But are we done yet? Not exactly, as we are missing an important piece: restarting the service when the configuration changes, in order to actually apply the configuration.

Restart the service (when needed!)

You can spot it is a bit different compared to our other methods: in this case it is not a state but an action we need to express. If we use an action method in our policies, the action will be executed at every agent run, which is not what we want here.

The technique editor has a concept of conditions. What are conditions?

A condition is represented by a string, and can be either defined or not. The conditions express what the current execution environment is:

  • We are on a Debian 9 system

  • The state of the nginx package is correct

  • The content of the configuration file has just been modified

  • etc.

We can use conditions to limit the evaluation of a method to a specific context, for example only on debian 9 or only when a given file has been modified by the agent.

This allows:

  • using actions (like service restart) by limiting them to a specific context

  • writing generic policies compatible with different operating systems, by having specific parts for each

In our case, we need to only execute our service restart when the configuration file method actually modifies the file. We need to use a condition that will be defined in this precise case.

Every method will define a result condition that is one of the conditions displayed in method details:

Install nginx package conditions

It can be:

  • Success: When the state was already compliant

  • Repaired: When the state has been modofied by the agent to become compliant

  • Error: When the expected state could not be reached

In our case the condition will be the modification of the configuration file, so the condition will be the repaired condition of out templating method, i.e. file_from_string_mustache__etc_nginx_conf_d_demo_conf_repaired (you should use copy/paste in most result condition definitions).

Conditions can be combinated using boolean operators:

  • ! for not

  • | for or

  • . for and

  • ( and ) for grouping

Now your service restart will only be executed when necessary.

When you start using conditions, be careful to only use them when necessary. For example, we could imagine only deploying our website when installing the package.

This is less reliable as we would stop checking for this symlink, and always consider it ok.

In short: Checking configuration is cheap, only add conditions when strictly necessary.


We now have defined a complete applicative example, which probably looks like what you would do with a real system. In the following sections, we will apply it to a machine and check the results.