Learning puppet chapter 2

LEARNING
PUPPET
- 02
SlideBook (Book on slides)
Inspired by Slidedoc -
http://www.duarte.com
-Vishal Biyani
www.vishalbiyani.com

Puppet
languagelanguage
for
“No Ruby”
person!

www.vishalbiyani.comLearning Puppet
Puppet
language
Puppet is a DSL based on Ruby – an object oriented programming
language. So do you need to know Ruby in and out to work with
Puppet ? Not necessarily!The whole idea of this chapter is to
introduce enough puppet & ruby to work with it.
Puppet is declarative which has two implications
• You define the final desired state not how to achieve it
• The sequence in which tasks are carried out is not guaranteed unless you
explicitly specify it.
You don’t have to master all of the concepts right here – but do
spend some time understanding syntax in this chapter.You can
3|
spend some time understanding syntax in this chapter.You can
always come back to understand any specific part which might
not be clear.
As with any language – there are sometimes multiple ways to do the
same thing and this is true for Puppet/Ruby too.
Lastly covering an entire language in such short span is a challenge.
So do refer to guides & references:
Some useful links:
• https://docs.chef.io/ruby.html provides ruby
basics needed for Chef framework. A good quick
introduction
• Similar quick guide from Puppet documentation:
http://docs.puppetlabs.com/puppet/3.7/reference
/lang_visual_index.html
• For the super curious once for quickly learning
Ruby there is no better source than
https://rubymonk.com/ that I know of!
Some language features are best explained within a
context – with a use case or a problem. Hence those
parts of puppet language have been intentionally
skipped here and will be covered as we go.

Variables &
Expressions
Lot of times we also need to resolve variables within a string for ex:
Variables can be declared with names starting
with $ and value can be a literal value or a
function/expression which resolves to a value.
1 $name = "vishal"
2 $number = 10+2
As long as the value that a variable resolves to is
compatible with data type of intended target, it
can be used as part of expression, function or
attribute!You will also see variables in following
6 $f_name = "${conf_dir}/.conf"
3 $users = ['us1','us2']
4 $users += ['us3']
Here we are resolving variable named conf_dir within a string – called
interpolation. Similarly addition of variables of same type is easy:
Now coming to one of most important features of variables – you can
assign a variable only once in a given scope. In a new scope you can assign a
new value to same variable, for example below the $name is reassigned but
in a new scaope – scope of the class testClass:
1 $name = "Default Name"
2 class testClass {
3 $name = "Test Class Name"
4 }
4|
attribute!You will also see variables in following
form:
4 $tomcat::config::tc_port
What does this mean? It means a variable called
“tc_port” from class tomcat::config. We will see
the classes & modules shortly.You can access
variables from parent scopes but not from child
scope (And we will scope in detail soon). For
example you can access a top level variable from
class level scope but not other way around. Also
there is a special notation for variables at top
scope – the scope is blank. In following example
we are getting “osfamily” from top scope
4 $::osfamily
4 }
Lastly – a variable can not be resolved unless it is defined.Which means
unlike rest of Puppet – order matters while defining variables.
Expressions
Most expressions in puppet are usual suspects of any programming
language like == (Equality), != (Non Equality), ,< , > , <= , >= , and, or, ! , +, -,
/, *, % etc. Couple of them that need some attention are not found in some
languages are:
=~ Left hand is a string & right hand is a regex – if match is found
then returns true
!~ Left hand is a string and right is a regex – if match is found
returns false
in Left hand operand should be string and right hand could be
string, array, hash. It checks if string on left side is “in” the right
side variable. Returns true if found

Of Conditions
We have quite a few friends in conditional
statements in Puppet – if, unless, case &
selectors.This part is quite easy so let’s run over it
quick!
Our if, elseif & else block is dead simple
1 if $osfamily == 'linux'{
2 # Do something
3 }
4 elseif $osfamily == 'windows'{
5 # Do something else
6 }
7 else
8 # Do something horrible here
But if we use regular expression in if block – there
Unless is exact opposite of if – but without additional else & regex magic:
1 unless $osfamily == 'linux'{
2 # Only for Linux family things
3 }
1 case $osfamily {
2 'Linux' : {include role::linux}
3 'Windows': {include role::windows}
4 default : {include role::default}
5 }
Case statement is also similar to other languages, almost like having
multiple if conditions & a default condition which will be executed if none of
previous conditions meet. And oh they also support regex capture variables if
you are using a regular expression to evaluate the case expression!
Finally the selectors!They are a special case of case conditional expression
and differ in behavior. In case statement we execute a piece of code once the
5|
But if we use regular expression in if block – there
is some black magic happening. If you look at
regular expression below – it is expecting string
some and varying digits after it – like some01,
some02 etc. Now these might be your hostnames
and would be good to get them inside the
execution block – as shown in comments within
the block.They have a name: “Regex Capture
variables”
10 if $hostname =~ /^some(d+)./ {
11 # Magic happens here with $0 and $1
12 # If some2 is passed then $1 has value
of 2
13 # and $0 has value of some2 - entire
match
14 }
and differ in behavior. In case statement we execute a piece of code once the
condition in a block is met, but in case of selector statement when the
expression evaluates to true – a value is returned.This value can be assigned
to a variable and then used later in flow. Imagine a situation where your
windows user name and linux username for certain operation is different –
and you want to use a selector to determine the user name dynamically
based on OS as shown in example below:
1 $user_name = $osfamily ? {
2 'windows' => 'win_admin',
3 'linux' => 'root_user',
4 default => 'root',
5 }
In above code – if $osfamily is windows then the username is different and if
it is linux then it is a different user.

Functions & Classes
Functions when it comes to puppet have a special
meaning.They are already defined pieces of code
which you can use to get things done. But more
importantly – functions are evaluated at
compilation time.Which means when your
manifest convert to catalog – the functions either
return their values or modify catalog based on
function!We have already used a function on
previous slide – which includes roles:
A class is a logical grouping of resources which can be accessed by class
name.There are two parts to it- first is you define the class and secondly you
call (Called declare) it.The execution happens only when you make a call.
Let’s first see how to define classes:
2 'Linux' : {include role::linux}
Another function which returns a value of true or
false based on input:
1 class test {
2 # Resource declarations here
3 }
4 class testParam ($name = 'default_value') {
5 # You can use the $name in any resource declaration
6 }
In first example above we defined a class which does not take any
parameters.Second one takes a parameter called “name” and if the caller
does not provide any value then “default_value” is chosen.This is just to
safeguard in case no value is provided & is a good practice to have a default.
The resources inside class are defined as you would do normally and they
6|
false based on input:
1 str2bool($is_virtual)
Functions have following characteristics:
• Rvalues type of function retuen some value (Like
str2bool)
•Statements kind of function will do some work but
wont return anything (Like include)
•Functions can take arguments – based on their
definition of what kind of & how many arguments
they take.
•Functions are run at compilation time and hence
on master – so any agent specific data/code should
be done using functions. Instead use
resource/custom facts (Which we will see soon)!
The resources inside class are defined as you would do normally and they
are belong/related to class.
The classes should be in files with name of class & ideally should be in
manifests directory within a module.You can of course place classes within
main manifest or manifests which are imported into main manifest, but for
now we won’t look at those cases.
Puppet also has concept of class inheritance but it should be used sparingly.
We will see the concept and when to use it in later chapters including how to
override parent resource attributes

Declaring Classes
Before we deep dive into declaring/calling classes
let’s understand a fundamental difference due
the way this evolved with Puppet:
•Include-like way:This was introduced in 3.x &
later versions.You can declare a class any number
of times but it will be called only once. It uses a
combination of external data & defaults to set
values of parameters. (Good practice of
externalizing configuration data from code). It is
suggested to use this way.
•Resource-like way:This is old way – in which
Let’s say you want to include a class into another one, this can be done
using “require” funcion which BTW is different than require used for
ordering.The require’s behavior is include like and you can pass list
separated by comma, array etc.
There is third include like way of declaring classes which uses hiera but we
will see that when we talk about hiera!
Now let’s look at resource like way of declaring classes which is not
suggested after version puppet 3. It is as good as declaring any other
resource:
6 define nix::tomcat(){
7 require java
8 }
10 class {‘nix::tomcat’}
7|
•Resource-like way:This is old way – in which
you declare a class the way you declare a
resource.You MUST DECLARE ONLY ONCE!You
can override parameters else they will be taken
from external data or lastly form defaults!
Now let’s see in how many different ways we can
declare classes.
Include: this is of course include like way to
declare classes and can take many forms
There is another way to assign classes to some nodes is via something called
ENC - External Node Classifier!Why would you do that?
External node classification is process of defining which node belong to
which classes. Doing it separate from automation tool provides loose
coupling & separation of concerns. ENC can be an external script or a
system which when called gives data of which nodes need to be applied
which classes and the automation tool accordingly applies the same. For
example many organizations store the node data in LDAP and in such case
you will fetch data from LDAP and then run puppet based on which nodes
needs which class.This is a topic which needs more detailed treatment and
we will not go into much details on this one at the moment.
1 include nix::tomcat
2 include nix::tomcat, nginx
3 clas_list = ['nix::tomcat','nginx']
4 include clas_list
10 class {‘nix::tomcat’}

Resource & types
Some more ground rules for resources:
•You can not declare the same resource twice! Imagine if in one place you
said file should be present and other place absent. Since puppet is declarative
language – it has to have a clear state defined. Although you can add
attributes to already defined resource for example:
1 file { "/home/vagrant/testFile"
2 ensure => present,
3 owner => 'root',
4 group => 'root',
5 mode => '0400',
6 }
Lifecycle of a resource:
When a resource declaration is executed, following things happen:
• The resource’s current state is read.
• If there is a different in current state vs. desired state then the change the
resource to desired state.
• The change in state is logged as event. This event will appear in log files
& reports.
The title of type in this case is
Every resource has a type – in our example below
it is file. It can be a service you want to manage or
a package that needs installed etc.
8|
attributes to already defined resource for example:
1 file { '/home/vagrant/testFile'
3 }
4 File['/home/vagrant/testFile']{
5 owner => 'root',
6 group => 'root',
7 mode => '0400',
8 }
The title of type in this case is
‘/home/vagrant/testFile’ & rest of the details are
”attributes” with values (Equivalent to
parameters in context of resources). Now the
“ensure” attribute will ensure that the file is
“present” – meaning it exists.The value of ensure
might vary from type to type – for example for a
service you might say “running” . So we are
defining the desired state of the a resource. How
to achieve it is what puppet does it for us on
multiple platforms. So in short a resource has
primarily three things – type, title & one or more
attributes. Attributes also might vary form type
to type.
•The order in which you declare resources in file does not matter. Meaning
during execution the order might change for execution optimization. So if you
want to introduce sequence in resource execution – you will have to use
relationships, which we will see shortly.

Objects & Scope 1/2 Now in real world scenarios when you configure nodes/machine – one
node might be Database server while another might be a Application
Server and so on. So you need to define what is a node’s identity! (Also
called node classification). In Puppet we can define a node’s identity in a
file /etc/puppet/manifests/site.pp – and this is our next level of structure.
Let’s say you have a Database node – you might want to configure OS
level resources first – which will be common across all type of nodes. After
that you might apply database specific resource definitions.These two
should be ideally in separate modules. So our node definition looks like:
Remember that everything in Puppet is
resource? But as we add more resources – there
is a need of a structure to manage them. It’s
almost like game of lego – you first get smaller
blocks and combine them to make bigger
structures.
We can combine related resources together to
create a class. And we can combine smaller
classes to create bigger classes. For example we
50 #/etc/puppet/manifests/site.pp
51 node 'database.mycompany.com' {
52 include os_common
53 include database
9|
classes to create bigger classes. For example we
might have a class for managing a file:
42 class base::home {
43 file { '/home/vagrant':
44 owner => 'root',
45 group => 'root',
46 mode => '0644',
47 }
48 }
Next we combine classes & configuration data to
create modules.These modules & classes can be
called from outside. Again we won’t go to depths
of module right away but consider that as next
logical level of structure.
53 include database
54 }
Here the modules os_common & and database is applied on a node
named “database.mycompany.com”
And now we have last structural element – “top”! It will be more clear
when we talk about scope in next slide but for now it is a level which
applies to all nodes in your infrastructure.

Objects & Scope 2/2
What is scope? A way to isolate & limit reach of things. In case of puppet
specifically scope applies to variables & resource defaults. Let’s checkout various
scope levels and their interplay:
TopScope: is what you define at top most level and is accessible everywhere!
Node Scope:Things that you define in your node definition exist at node scope.
Code you define at node scope is available to classes included in that node but
not at top scope.
Local scope is something local to a class, resource declaration or a method etc.
Overriding is the means to override a variable's value form parent scope. Let’s
say we define “message” at node scope and need to override in a class:
We saw four levels in last slide: Resource, Class,
Node & top. Module is just a structural element
which encompasses classes and is not a scope level
really.Scopes form a hierarchy – for example you
can access top scope code at node scope but not
other way around. Basically a parent scope
code/variable can be accessed at child level but child
level can not access things at parent level. Below
picture illustrates the scopes at various levels.
topTop Scope
1 node example.com {
2 $message = "This is node"
3 include
4 }
10|
resource
resource
resource
resource
resource
classclass
module
class filesconfig
Node definition
Local
Scopes
Node Scope
4 }
5 class test_class {
6 $message = "This is class overriding node"
7 notify{"Test message = $message":}
8 }
What will get printed above? Message from class due to overriding.
The scopes of top & class get names – i.e. they can be references using a
notation like $::variable etc. But node & local scopes don’t get names hence
they can not be referenced. Finally the variables which are not in your scope and
have name (i.e. do not belong to node/local) can be accessed with their full
names. For example let’s say the class tomcat has a parameter named port, we
can call it with it’s fully qualified name:
1 include tomcat::config
2 $port = $tomcat::config::port_num

Relations & Ordering
We saw that the sequence of execution of resources
can be different than what we see in manifest. But
there are times when you need to enforce order –
and that’s where ordering attributes help us.There
are four attributes – or metaparameters which
enable ordering:
Attribute Effect
before Execute current resource before and
Let’s say we want to ensure package “ssh” is installed first and then the file
“/etc/ssh/ssh_config” is configured. So there are two ways to do it (With only
relevant attributes:
10 file {'/etc/ssh/ssh_config'
12 require => Package[ssh]
13 }
14 package {'ssh'
15 ensure => installed,
16 }
20 }
21 package {'ssh'
22 ensure => installed,
23 before => File['/etc/ssh/ssh_config']
24 }
Similarly let’s say we want the service SSH to be running only after the file
“/etc/ssh/ssh_config” is configured. In addition the service SSH should be
11|
The notation -> (ordering) and ~> (Ordering + Refresh) are equivalent to the four attributes above based on direction you use
them. You can use them or be more explicit and use the attribute names. These arrows are lovingly called “chaining arrows”
before Execute current resource before and
target resource later
require Execute current resource later and
target resource before
notify Execute current resource before,
target resource later & refresh target
resource of current one changes
subscribe Execute target resource before, then
execute current resource, and refresh
current if target has changed
As you have noticed – before and require are same
things but depends on where you declare it. Ditto
with notify & subscribe but with a refresh included.
Quite confusing? – let’s look at some example.
“/etc/ssh/ssh_config” is configured. In addition the service SSH should be
restarted every time the file changes. Again two ways we can configure it:
28 notify => Service['ssh'],
29 }
30 service {'ssh'
31 ensure => running,
32 }
36 }
37 service {'ssh'
38 ensure => running,
39 subscribe => File['/etc/ssh/ssh_config']
40 }
• You can refer an array of resources in value for attributes. For example you can
restart SSH and some other service as file changes.
• There are more complex scenarios where you might want to use “resource
collectors” – but more on that later ☺

What did we learn?
Although we covered quite a few
things, we have omitted many
details and certain concepts
altogether in this chapter.This is
intentional and we will pick up
those concepts with examples so
that they make lot more sense
than bunch of theory concepts.
Let’s start building stuff now!
This chapter seems like lot of
theory but as stated in
introduction – do a sanity round
and come back to this chapter
every time you come across a
certain concept and need help
implementing it.We have
covered quite a few things and
we will start using them in
12|
Let’s start building stuff now!we will start using them in
upcoming chapters.

Learning puppet chapter 2

Recommended

More Related Content

What's hot (20)

Viewers also liked (18)

Similar to Learning puppet chapter 2 (20)

More from Vishal Biyani (15)

Recently uploaded (20)

Learning puppet chapter 2