Introduction to Chef: Automate Your Infrastructure by Modeling It In Code

Automate Your Server Conﬁguration!
by Modeling It In Code
Josh Padnick
Desert Code Camp
April 5, 2014

Today’s 5-Course Meal
• What is Chef and what does it solve?
• Key Chef concepts
• Where/How you use it
• Tips & tricks
• Learning more

Our Goal Today:
You go home tonight and deploy your
ﬁrst Chef cookbook in 1 - 2 hours.

• Founder & CIO of Omedix
• Special interest in Healthcare IT
• 10+ years doing web app dev
• Strong preference for open source & Java ecosystem
Josh Padnick
These slides are posted on http://joshpadnick.com

Some Problems from the
Server Management World

“The Magic Server”
The code doesn't work on any server
except the magic server. We're afraid to touch it.

“Just Clone Another Instance”
You can’t setup a dev machine quickly, but you're
virtualizing and at least you can clone it.

“We have AMIs! But we need to change them :(“
You can clone a server, but your snapshot (AMI) needs
to be changed...so you have to rebuild the AMI.

“I have a simple script to deploy!
to 10 different servers”
The script is small and simple, but deploying it to 10 servers
is painful, so updates to this script will not be permitted!

Documentation is time-consuming and always out of date.
But we still need to ramp up the new guy.
“Documentation is critical…but out of date.”

"Deploying a new server takes too long"
!
"This server here is broken so let's spend hours
restoring it because launching a new server is too hard."
!
"Changing the deployment process will take time we
don't have"
!

Are you actively managing more than
a few servers on an ongoing basis?
Welcome to!
SERVER MANAGEMENT HELL

What is the underlying issue here?
The real problem is STATE.

But wait, haven’t we dealt
with state before?
Only since 1937 when Alan Turing invented the state machine.

What does the Chef solution look like?

The Servers
Management Server
This guy is the authority on what
state each server should be in

The Servers
Management Server Local Workstation

The Servers
Each of these guys installs an
agent called the CHEF CLIENT

The Servers

The Servers
The DevOps engineer pushes all
instructions to the Management Server

Now let’s use all the Chef lingo.

Nodes
Chef Server Knife
Write code that describes the
state of a node and deploy it
to Chef Server

Write code that describes the
state of a node and deploy it
to Chef Server
Has the ofﬁcial record of what
each server’s state SHOULD be
Nodes
Chef Server Knife

Each node periodically polls the
Chef Server asking for “update
state” instructions
Has the ofﬁcial record of what
each server’s state SHOULD be
Nodes
Chef Server Knife

Each node updates its state!
based on instructions from the
management server.
Nodes
Chef Server Knife

Nodes
Chef Server Knife
Each Node’s state has been
updated!

Did someone say we’re
writing code?
• Write it in Ruby 2.x
• Chef gives us a Ruby DSL specially for declaring server state
• Version-control it with anything, but standard is git

We’ll go into more Chef detail later.
Let’s pan out to the 50,000 foot view again.

Chef is considered a
Conﬁguration Management!
Solution

What is Conﬁguration Management?
• A tool used to manage server conﬁguration 
with automation.
• Born out of the need for major websites
(Amazon, Facebook, Yahoo) to manage huge
numbers of servers.

Who Created Chef?
Jesse Robbins Adam Jacob
Managed lots of Amazon.com servers
Real-life ﬁre ﬁghter!
Built infrastructure for 15 companies
Kept seeing the same patterns!

Alternatives to Chef
• Ansible
• Saltstack
• Puppet

DevOps
• Chef merges the worlds of Development (Dev)
and SysAdmin (Ops).
• Dev: build software, version control, automated testing
• Ops: provisioning servers, maintaining servers, monitoring

Let’s talk more about key
concepts in Chef.

• The primary currency in Chef is the Cookbook.
• That’s the set of code and ﬁles we use to tell a
server what state to update itself to.

• Sample cookbooks
• MySQL
• Nginx
• Install Java
• Update crontab
• Update timezone
• Add SSH certiﬁcates to authorized_certiﬁcates

• You can roll your own
• But for almost every need, just search GitHub
• google “github java cookbook” and choose the best one
• Opscode has “Community Cookbooks at http://
community.opscode.com/cookbooks but I ﬁnd it
out of date and incomplete.

Sample Cookbook
• Recipes are individual sets of
instructions to be executed.
• Recipes read values from
Attributes for things like:!
• Passwords
• Filepaths
• Usernames
• Conﬁguration options for applications

Attributes
• How attributes get applied to a recipe are behind
much of the perceived complexity of Chef.
• We’ll cover this in more detail later.

Let’s see how a cookbook ultimately
gets executed on a node.

Nodes
Chef Server Knife
At your workstation, download or
write a new cookbook. Let’s call
it “Sample”.

Nodes
Chef Server Knife
Upload your cookbook to the
Chef Server using KNIFE.

Nodes
Chef Server Knife
Now has the latest version of
the “Sample” cookbook.

Nodes
Chef Server Knife
Use KNIFE to tell CHEF SERVER
the RUN-LIST for a speciﬁc node.

Nodes
Chef Server Knife
Now knows that Node 32 should
run the “Sample” cookbook.

Nodes
Chef Server Knife
SSH directly into an individual node
(“Node 32”) and run “sudo chef-
client” to CONVERGE the node.

Nodes
Chef Server Knife
Node 32 uses chef-client to
contact Chef Server. It get its
updated run-list, and executes
the run list.

Nodes
Chef Server Knife
Node 32’s run-list says to run the
Sample Cookbook. It runs the
latest version of Sample.

Nodes
Chef Server Knife
Node 32 has now executed all the
instructions in the latest version of
the “Sample” cookbook

Wait, we have to manually log into
each node to update it?
• No! In production, we use Roles to specify a run-list.
• This way, we only update the Chef Server. Individual
nodes poll the Chef Server every X minutes to check
for updates.

• Traditional Chef Server
• AWS OpsWorks
• Clould Management (Scalr & RightScale)
• Vagrant

Traditional Chef Server
Nodes
Chef Server Knife

Traditional Chef Server
• Chef Server comes in Open Source or Enterprise
Edition
• Chef Server is “on premise” or hosted by Opsworks

Enterprise Chef Pricing
http://www.getchef.com/enterprise-chef/#plans-pricing

AWS OpsWorks
Local Workstation

Different Flavors of Chef-Client
• Server Mode
• Chef Solo Mode
• Local Mode

Cloud Management Services
(e.g. Scalr & RightScale)
Nodes
Local Workstation

• It’s a command-line interface to virtual machine
software like VirtualBox or VMWAre.

• You can combine Chef’s cookbooks, VirtualBox (a
free VM provider), and Vagrant to run chef
cookbooks directly on local VM’s!
• The same cookbooks that deﬁne your infrastructure
can now deﬁne your local dev environment.

• Get the latest cookbooks on
your local machine
• Run a bunch of VM’s with
VirtualBox and Vagrant
• Update the VM’s with Chef
cookbooks
Local Workstation

http://docs.opscode.com/chef_overview.html

Remaining Concepts We’ll Talk About
• Roles
• Environments
• Data Bags
• Ohai
• Source Control & Chef

Remaining Concepts We Won’t Cover
• Resources
• Providers
• LWRPs

All those remaining concepts we’ll cover fall into two buckets:
1. Attributes
2. Run-lists

Attributes
• Can be set in multiple places

Ohai
• It’s a program that runs on each node and supplies
attribute info speciﬁc to that node.
• Examples
• What OS the node is running
• How much hard drive space
• How much memory is available
• Linux kernel version

Roles
• The Run-List is usually the same for all servers at the
same “layer” in the stack and different across layers.
Runlist[Apt, Sample, Apache]
Runlist[Apt, Java, Tomcat, JBoss]
Runlist[Apt, Postgres, Newrelic]
“web” Role
“app” Role
“db” Role

Roles
• Roles can also specify attributes that OVERRIDE the
“default” attributes set in the Recipe.
• When a new node is created (“bootstrapped”) it is
best practice to explicitly identify which Role it
belongs to.
• Roles are declared as a simple JSON ﬁle and
uploaded to Chef Server using Knife.

Nodes
Chef Server Knife
Use KNIFE to bootstrap a new
NODE with the “web” ROLE.

Nodes
Chef Server Knife
A new NODE exists that is
assigned to the “web” role

Nodes
Chef Server Knife
Use KNIFE to update the “web”
ROLE to add a new RECIPE to
the RUN-LIST.

Nodes
Chef Server Knife
The “web” role has been
updated.

Nodes
Chef Server Knife
New Node polls the CHEF
SERVER every X minutes to look
for updates.

Nodes
Chef Server Knife
The “web” ROLE now includes a
new RECIPE so download this
COOKBOOK and run the RECIPE

Nodes
Chef Server Knife
All nodes in the “web” role have
“CONVERGED.”

Environments
• We typically have a PROD and DEV. Maybe QA and
STAGING, or others.
• Environments are just another label to assign to a
node so that it gets the right attributes.
• Just like Roles! But with a different name and intent.

Data Bags
• Data Bags are a global source of attributes that any
recipe can call upon.
• They work great for global attributes
• Not so great for secrets like passwords. More on this
later.

Source Control & Chef
• Your local chef repo should be cloned from https://
github.com/opscode/chef-repo.
• Then commit it to your own Git repo so you can
version-control changes to Cookbooks, Roles,
Environments, Data Bags, etc.
• Now you version-control your infrastructure just like
your code itself!

Attribute Hierarchy
• Hard to learn just by looking. Learn by doing.

Tips & Tricks
Also hard to remember. Intended for reference!

Storing Secrets in Your Infrastructure
• This is a very hard problem! Let’s look at some options.
• Option 1: The ofﬁcial Chef solution is encrypted data bags. But the main
problem is all nodes and the Chef server share the same symmetric
encryption key :(. So how do we securely transport and protect that key?
• Option 2: Nordstrom uses Chef and created something called Chef Vault
to replace the symmetric encryption key of encrypted data bags with
public key infrastructure. Works well, but creates the “chicken and egg”
problem where a server can’t register itself with chef-vault until it’s
bootstrapped, but needs secrets from chef-vault to bootstrap itself.
• BEST OPTION for AWS! Option 3: Use Citadel. (https://github.com/
balanced-cookbooks/citadel). Store all your secrets in an S3 bucket.
Lock down S3 with AWS IAM Users. Assign each EC2 instance (node) to
an IAM Role which automatically grants access to that instance to the S3
buckets we specify. No keys to manage b/c Amazon does it for us!

Open Source Chef Server Tips
• Follow the instructions at http://docs.opscode.com/install_server.html
• Go to http://www.getchef.com/chef/install/ to get the URL for the file download
• To setup the Fully Qualified Domain Name (i.e. hostname) for Ubuntu, do this:
• Setup a DNS name for the server (chef.mybiz.com)
• sudo vim /etc/hostname and enter the hostname to handle server reboots
• sudo hostname chef.mybiz.com to change the hostname for the current session  
• Immediately setup a user/pass for yourself so that admin remains a "root" account.
• For AWS, a m1.small instance is sufficient for now.
• You will need backup and monitoring for this server.
• See http://www.getchef.com/blog/2013/03/11/chef-11-server-up-and-running/ for more info.
• You can bootstrap Chef Server with Chef Solo! See https://github.com/opscode-cookbooks/
chef-server

Tips for Setting Up Knife
• When you run knife configure --initial use your local paths for the admin.pem and the chef-
validator.pem
• At some point, you'll need to download files from /etc/chef-server folder on the Chef Server in order to
get Knife up and running.
• You'll need to modify your knife.rb file (e.g. to point to your cookbooks path) to get things working
right. Errors caused by this are not well documented. Here's my knife.rb:
Joshs-MacBook-Pro:.chef josh$ vim knife.rb !
!
log_level :info!
log_location STDOUT!
node_name 'josh'!
client_key '/Users/josh/.chef/josh.pem'!
validation_client_name 'chef-validator'!
validation_key '/repos/chef-repo/.chef/chef-validator.pem'!
chef_server_url 'https://chef.projname.mybiz.com'!
syntax_check_cache_path '/Users/josh/.chef/syntax_check_cache'!
cookbook_path '/repos/chef-repo/cookbooks'!
knife[:aws_access_key_id] = "Your AWS Access Key ID"!
knife[:aws_secret_access_key] = "Your AWS Secret Access Key"!
knife[:region] = "us-west-2"!
knife[:vpc_id] = "vpc-XXXXXXX"

Writing Cookbooks
• Having a fast feedback loop is key. Also don’t want
to rack up AWS costs.
• Ideal environment for writing cookbooks is EC2 tiny
instances with a Chef Server or use Chef Solo with
Vagrant.
• Use test-kitchen to help manage your Vagrant
environment.
• User berkshelf to help manage cookbook
dependencies (if it’s getting out of hand)

Helpful Ruby Tips
• I knew zero Ruby when I started and got by ﬁne.
Here are the only confusing things I encountered:
• In Ruby, :stringLiteral is called a “symbol” and is equivalent to
“stringLiteral”. See http://www.reactive.io/tips/2009/01/11/the-difference-
between-ruby-symbols-and-strings/
• A “heredoc” refers simply to a multiline string and is begun by <<EOH
and ended with EOH on a newline.
• Everything else is pretty straightforward.

Getting Started &
Learning More

Start with These Milestones
1. Setup Chef Server (hosted or on-premise)
2. Setup Knife
3. Setup Vagrant environment
4. Write (or download) Cookbooks!

Where to Learn
• Start here: https://learnchef.opscode.com. Screencasts are a
perfect place to begin.
• Then go to http://docs.opscode.com (walkthrough) for more info.
• Check out http://gettingstartedwithchef.com as another
reference.
• Study other people’s cookbooks to get ideas. The postgresql
cookbook is very well done.
• #chef on IRC was very helpful for me.
• The initial learning curve is somewhat steep, but it quickly
becomes fun!

El Fin
 
Now go cook something delicious!

Introduction to Chef: Automate Your Infrastructure by Modeling It In Code

Recommended

More Related Content

What's hot (20)

Similar to Introduction to Chef: Automate Your Infrastructure by Modeling It In Code (20)

Recently uploaded (20)

Introduction to Chef: Automate Your Infrastructure by Modeling It In Code