A VMware Administrator's Guide to Setting Up a DevOps Configuration Management Test Environment
Date: Apr 13, 2015
Before we dive headfirst into implementing DevOps tools, let’s examine how we can set up our test environments to adequately prepare ourselves. We’ll also take a look at how we can start treating our infrastructure instructions just like a developer’s source code.
Topics covered in this chapter include the following:
- Environment provisioning with AutoLab
- Environment provisioning with Vagrant
- Creating images with Packer
- Managing source code
- Git source code control
Environment Provisioning with AutoLab
AutoLab is a vSphere test lab provisioning system developed by Alastair Cooke and Nick Marshall, with contributions from others in the VMware community. AutoLab has grown so popular that at least one cloud provider (Bare Metal Cloud) allows you to provision AutoLab installations on its servers. As of this writing, AutoLab supports the latest release of vSphere, which will enable you to run a virtual lab that includes VSAN support. If you’re unfamiliar with the tool, head over to http://www.labguides.com/ to check out the AutoLab link on the home page. After filling out the registration form, you will have access to the template virtual machines (VMs) to get your own AutoLab setup started. The team has a helpful selection of how-to videos on the AutoLab YouTube channel for anyone needing extra assistance.
Environment Provisioning with Vagrant
Vagrant is an environment provisioning system created by Mitchell Hashimoto and supported by his company, HashiCorp. Vagrant can help you quickly bring up VMs according to a pattern defined in a template file known as a Vagrantfile. Vagrant can run on Windows, Linux, and OS X (Mac) operating systems and supports popular desktop hypervisors such as VMware Workstation Professional, VMware Fusion Professional, and VirtualBox. Cloud providers such as Rackspace and Amazon Web Services can be used as well for your test environment. The Vagrant examples in this book are based on the VMware Fusion plugin, but the examples we provide can, with a few modifications, be used for other hypervisors and for cloud platforms. If you will be following along with the examples in this chapter, make sure to create a new directory for each Vagrantfile that you create.
After installing Vagrant and the VMware Fusion or Workstation plugin, you need to find a Vagrant box to use with the system. A Vagrant box can be thought of as a VM template: a preinstalled operating system instance that can be modified according to the settings that you specify in your Vagrantfile. Vagrant boxes are minimal installations of your desired operating system with some boxes not being more than 300 MB. The idea is to present a bare-bones operating system that is completely configured by your automation tool of choice.
Some box creators opt to include the binaries for popular Vagrant-supported provisioners like Puppet, Chef, and so on, but other box creators do not, and users need to use the shell provisioner to deploy their favorite configuration management solution before it can be used. The box creation process is beyond the scope of this book. However, if you would like to develop your own boxes, there are tools like veewee and Packer (discussed later in this chapter) available that you can try out.
In previous versions of Vagrant, you had to specify the URL of the box file that you want to use when you initialized your vagrant environment:
vagrant init http://files.vagrantup.com/precise64_vmware.box
If the box file is located on your computer, you can specify the full path to the file instead of a URL.
Starting with Vagrant version 1.5, HashiCorp introduced the Atlas system (formerly known as Vagrant Cloud), an online repository that Vagrant will search for boxes if you use the account and box name of an image stored on its site:
vagrant init hashicorp/precise64
It is good to know both types of syntax because it will be necessary to use the old method for any boxes not hosted on Atlas. The online site is a great place to search for boxes for various operating systems instead of building your own.
The vagrant init command will automatically create a simple Vagrantfile that will reference the box that you specify. Listing 3-1 shows the default Vagrantfile that is generated with the command listed above.
Listing 3-1 Default Vagrantfile
# -*- mode: ruby -*- # vi: set ft=ruby : # Vagrantfile API/syntax version. Don't touch unless you know what you're doing! VAGRANTFILE_API_VERSION = "2" Vagrant.configure(VAGRANTFILE_API_VERSION) do |config| config.vm.box = "hashicorp/precise64" end
You’ll notice that to save space I remove the comments that automatically get generated when you initialize your Vagrantfile. However, if you look in the comments, you’ll see some helpful tips for using configuration management technology to make automated changes to your VM when it boots. As of today, the available options for provisioning your VM include basic shell scripts and configuration management tools such as Puppet, Chef, and Ansible. This is an immense value because your development and test environment can be stood up with the exact same settings that are used in your production deployments. This should cut down on the “well, it worked on my laptop” discussions that may go back and forth during a deployment mishap. Docker support was also added so that the provisioner could install the Docker daemon automatically and download the containers that you specify for use.
With your Vagrantfile in place, you can now boot your first test environment by using the following command:
vagrant up --provider=vmware_fusion
You can now log in to the VM using the following command:
vagrant ssh
Take a look at a folder in your VM called /vagrant. You’ll see that it contains your Vagrantfile! It’s a shared folder that is automatically created for the VM so that you can easily transfer files to and from your desktop without having to use SCP, FTP, and so on.
If you examine the operating system resources, you’ll notice that you have one vCPU and 512 MB of RAM. This may not be sufficient for the application that you want to run. So, we will take a look at how to modify the resources allocated to your Vagrant VM.
First, let’s destroy this VM so that we can move on with the other configuration options. You can do this exiting the VM and then using the following command:
vagrant destroy
Vagrant will ask you to confirm that you really want to destroy this VM. Alternatively, you can use the -f option to skip that confirmation.
Listing 3-2 shows that Vagrant can modify the VM’s VMX file to make the changes that we need. We use the config.vm.provider block of code to achieve this. By the way, the memsize attribute’s units are megabytes. Notice that we are creating an object named v enclosed in vertical lines to change settings just for this VM. This object name has local scope only to this config.vm.provider statement, and it can be used again when defining other VMs, as you’ll see in later examples. After executing vagrant up, the VM will be created with the desired attributes. At the time of this writing, the size and number of virtual disks cannot be controlled, but your Vagrant VMs will start with 40 GB of thin-provisioned storage.
Listing 3-2 Changing Default Vagrantfile
# -*- mode: ruby -*-
# vi: set ft=ruby :
# Vagrantfile API/syntax version. Don't touch unless you know what you're doing!
VAGRANTFILE_API_VERSION = "2"
Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
config.vm.box = "hashicorp/precise64"
config.vm.provider :vmware_fusion do |v|
v.vmx["memsize"] = 1024
v.vmx["numvcpus"] = 2
end
end
It is great that we can modify the VM’s resources. What about a more complex setup, like multiple VMs? Vagrant supports such a topology as well. Of course, make sure that you have sufficient CPU cores and RAM to support the topology that you want to use! Multi-VM setups would be useful for testing realistic deployments with a separate database server and front-end server, for example. Listing 3-3 shows an example of a multi-VM Vagrantfile setup.
Listing 3-3 Multimachine Vagrantfile
# -*- mode: ruby -*-
# vi: set ft=ruby :
# Vagrantfile API/syntax version. Don't touch unless you know what you're doing!
VAGRANTFILE_API_VERSION = "2"
Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
config.vm.define :first do |vm1|
vm1.vm.box = "hashicorp/precise64"
vm1.vm.hostname = "devops"
vm1.vm.provider :vmware_fusion do |v|
v.vmx["memsize"] = 1024
v.vmx["numvcpus"] = 2
end
end
config.vm.define :second do |vm2|
vm2.vm.box = "hashicorp/precise64"
vm2.vm.hostname = "vmware"
vm2.vm.provider :vmware_fusion do |v|
v.vmx["memsize"] = 1024
v.vmx["numvcpus"] = 2
end
end
end
The deployment utilizes multiple config.vm.define blocks of code: one for each VM that we are creating. :first and :second are labels that Vagrant will use to identify the two VMs when you run commands like vagrant status. These labels will also be used to connect to the VMs via Secure Shell (SSH)—for example, vagrant ssh first. If you’re familiar with Ruby, you’ll notice that these labels are Ruby symbols. The names in the enclosed pipe symbols (for example, |vm1|) denote the object whose information that vagrant is using to build and define your VM. The object name can be the same as the symbol (for example, first.vm.box...), but it doesn’t have to be.
Using this syntax can be a bit tedious when you want to deploy more than two VMs. Thankfully, because Vagrant is written in Ruby, you can use the language’s features such as lists, loops, and variables to optimize your Vagrantfile code. Listing 3-4 shows some optimization tips that I learned from Cody Bunch and Kevin Jackson in their OpenStack Cloud Computing book
Listing 3-4 Optimized Multimachine Vagrantfile
# -*- mode: ruby -*-
# vi: set ft=ruby :
servers = ['first','second']
Vagrant.configure("2") do |config|
config.vm.box = "hashicorp/precise64"
servers.each do |hostname|
config.vm.define "#{hostname}" do |box|
box.vm.hostname = "#{hostname}.devops.vmware.com"
box.vm.provider :vmware_fusion do |v|
v.vmx["memsize"] = 1024
v.vmx["numvcpus"] = 2
end
end
end
end
At the top of the file, I create a Ruby list called servers whose elements are the names of the VMs that I want to create. Then I use the Ruby list iterator called each to loop the execution of the VM definition for each element in the servers list. If we ever want to increase the number of VMs that are deployed, we just add more entries to the list. Not every VM needs to have the same set of resources, and we can use if statements within the box.vm.provider code block to be selective:
if hostname == "first" v.vmx["memsize"] = 3128 v.vmx["numvcpus"] = 4 elsif hostname == "second" v.vmx["memsize"] = 1024 end
There are many more features in Vagrant that we will not be covering in this book, but with just these simple commands, you can build the test environment setups that we will be using in this book. If you’d like to learn more about Vagrant, be sure to check out the Vagrant website (http://www.vagrantup.com) and Mitchell’s book, Vagrant: Up and Running.
Creating Images with Packer
Packer is another HashiCorp product that helps you to develop your own boxes for multiple platforms. Let’s say you wanted to develop a VM image for Workstation/Fusion and ESXi from the same base box. Packer makes that possible.
Packer uses a JavaScript Object Notation (JSON) file format for you to specify how your Vagrant box will be configured (disk size, memory, and so on), and it will perform the initial OS deployment on your behalf once you specify the relevant automation parameters (for example, Ubuntu preseed files).
Packer is not just useful for creating Vagrant boxes; its main purpose is to produce image files that are compatible with popular cloud provider formats (OpenStack, AWS, and so forth). However, Packer includes a builder capability to automatically output Vagrant boxes that are compatible with VMware Fusion/Workstation and VirtualBox. Just like with Vagrant, popular configuration management technologies like Puppet and Chef can be used to customize the image that is produced.
Although we will not be discussing Packer in depth, we wanted you to be aware of it if you would like to experiment on your own with building custom Vagrant boxes. You can find more information about Packer at http://www.packer.io. If you would like to see examples of Packer definition files that you can use to develop your own VMs, the Chef team maintains a repository called bento on their Github account. https://github.com/chef/bento
Managing Source Code
Source code management (SCM) is an essential element in a DevOps environment. Think about it: If you will be turning your infrastructure into code, it is important that there is a way to review any changes and go back to different versions of a file in case new changes introduce problems (for instance, periodic instabilities in the best case, outages in the worst case). Some might think the “easy” way would be to make multiple copies of a file each with a unique name (Vagrantfile1, Vagrantfile2, Vagrantfile01012015, and so on), but then you have to deal with the hassle of renaming the file when you want to use it and trying to remember what was different about all the files.
The various teams in the development organization are most likely using some SCM system already to manage their work (for example, software developers storing their source code, QA teams managing their test scripts). As you begin using SCM technologies, it would be worthwhile discussing best practices with these other groups.
There are many SCM solutions, including SVN, Mercurial, and so on. Git happens to be one of the more popular SCM systems in the DevOps community. So, for this book, we use Git.
Using Git
Git is a distributed version control system, which means that you make a local copy of the central repository instead of just checking out individual files. Local commits can be synchronized back to the central server so that there is consistency in the environment, and users can always pull the latest version of the source from the central repository. This architecture differs from traditional source code management systems, in which only the central server ever has a complete copy of the repository.
As you work through the examples in this book, we recommend using one of the freely available online Git repositories, such as Bitbucket, GitHub, and Gitorious, as your central location for storing your code. Each site has its own unique features. For example, BitBucket allows unlimited free private repositories. GitHub is the online repository system that this book’s authors used for their code. However, feel free to use whichever system meets your needs, as the methods by which you obtain code (clone/pull) and store code (push) are universal across any Git system.
Creating Your First Git Repository
First, install Git using your favorite package manager (for example, homebrew or macports on Mac OS X, apt-get on Ubuntu/Debian, yum on Red Hat/CentOS/Fedora). For Windows users, the popular online repositories like GitHub and BitBucket offer software clients that make it easy to interact with their online repository system. Alternatively, the http://git-scm.com site maintains a standalone install of the Git binary for Windows.
If you are using Linux or Mac OS X, you can open a terminal window to work with the following examples. Windows users must use the special shell that gets installed with whichever Git client that you use. The example syntax will be Linux/Mac-based, but the commands should be equivalent for the Windows platform.
Before we start writing any code, we need to set a couple global variables so that Git knows who we are. It’s not as critical for local copies of the repository, but when we start pushing our code to a remote server, it will be very critical. The two global variables are your email address and username:
git config --global user.email "you@example.com" git config --global user.name "Your Name"
As a matter of fact, if you try using Git and making your first commit without setting these variables, Git will prompt you to set them before you can continue.
If you followed along with the earlier Vagrant examples, you already have a directory of content that we can work with. Otherwise, create a new directory and create a text file in it. From here on out, make sure that you are in that directory on your command-line prompt.
First, let’s initialize this directory to have a Git repository:
git init
If you do a listing of your directory with the option to show hidden files (ls -a on Linux/Mac or dir /A:H on Windows), you’ll see that there is a hidden directory called .git. This directory contains your repository’s files and settings specific to this repository. These local settings are combined with the global settings that we set earlier, and we can confirm this by using the following command:
git config –l
If you want to see the state of the files in your directory (has the file been added to the repository? Are there any changes since the last command? and so on), you can type git status and see output similar to what is shown in Listing 3-5.
Listing 3-5 Git Repository Status
git-test $ git status
On branch master
Initial commit
Untracked files:
(use "git add <file>..." to include in what will be committed)
.vagrant/
Vagrantfile
nothing added to commit but untracked files present (use "git add" to track)
The last line is most important; it tells us that our files need to be tracked for the repository to manage it. This is important to take note of as putting files into the directory does not automatically get it tracked by the SCM tool. This feature prevents us from tracking junk files in the repository and wasting space.
Let’s tell Git to track our Vagrantfile:
git add Vagrantfile
However, the .vagrant/ directory is not essential to be tracked because it only contains temporary files that Vagrant uses to set up your VM. We can explicitly tell Git to ignore this directory by creating a .gitignore file. Use your favorite text editor and create your .gitignore file with a single entry:
.vagrant/
Alternatively, you could use a simple echo command to accomplish the same thing. (Windows users will need to use the special Git Shell binary that is included with their Git install for this to work properly.)
echo '.vagrant/' > .gitignore
If you run the git status command again, you’ll see that Git informs us about the .gitignore file as well. What gives? Remember that Git needs to be told what to do with any files or directories that the repository can see including the .gitignore file. Well, there are two ways to deal with your .gitignore file:
- Add the .gitignore file itself to the list of files and directories to ignore.
- Tell Git to track the .gitignore file as well.
I will use the second option so that anyone else who may use my repository will be able to ignore the appropriate files as well:
git add .gitignore
Now, if we check the status of the repository again, we should see output similar to what is shown in Listing 3-6.
Listing 3-6 Updated Repository Status
git-test $ git status
On branch master
Initial commit
Changes to be committed:
(use "git rm --cached <file>..." to unstage)
new file: .gitignore
new file: Vagrantfile
All the files in our directory are either ready to be committed or added to the .gitignore list of nonessential files and directories. So, all that’s left to do is to commit the repository changes:
git commit
A file editor will automatically be opened so that you can enter details about the files that you are committing. (By default, vi is used.) See Listing 3-7.
Listing 3-7 Your Commit Message
git-test $ git commit 1 This is my first commit. 2 # Please enter the commit message for your changes. Lines starting 3 # with '#' will be ignored, and an empty message aborts the commit. 4 # On branch master 5 # 6 # Initial commit 7 # 8 # Changes to be committed: 9 # new file: .gitignore 10 # new file: Vagrantfile 11 #
You must enter a message; otherwise, the commit will be canceled. If you are not familiar with vi, press I, type some text, press the Escape key, and then type :wq and press Enter. If you don’t want to deal with the text editor, you can use the short form of the git commit command with the -m option to enter your commit message on the same line:
git commit -m "This is my first commit."
If the commit is successful, you should see the following output:
[master (root-commit) d962cd6] This is my first commit. 2 files changed, 119 insertions(+) create mode 100644 .gitignore create mode 100644 Vagrantfile
Working with a Central Git Server (a.k.a. A Remote)
If you have opened an account on either GitHub, BitBucket, or whatever public Git repository site that you prefer, you will need to provide your computer’s SSH public key to the site so that it can verify who you are. Each site may have a different way of doing this. So, consult the documentation for the appropriate steps. For Windows users, this is typically handled for you automatically by installing the client software for the site. Mac and Linux users must generate an SSH public key by using the ssh-keygen command.
Once your SSH public key is properly configured on your remote, it’s time to create the repository on the remote site that you will be storing your files into: a process known as pushing. When you create your remote repository on the website, you should skip the automatic generation of the README file. After the repository is created, the site will provide you with a link that you can use to tell your local repository what is the location of your remote server, often labeled as origin.
In my setup, I gave the same name to my repository as I did to my local repository. This is optional, and the names can differ. I can use the git remote command with the link that GitHub gave me to update my local repository settings:
git remote add origin git@github.com:DevOpsForVMwareAdministrators/git
If I use the git config -l command, I will see new data about the location of my remote server:
remote.origin.url=git@github.com:DevOpsForVMwareAdministrators/git-test.git remote.origin.fetch=+refs/heads/*:refs/remotes/origin/*
I can now push the files from my local repository to my remote repository:
git push origin master
I should see output similar to what is shown in Listing 3-8.
Listing 3-8 Git Remote Push Results
git-test $ git push origin master Warning: Permanently added the RSA host key for IP address '196.30.252.129' to the list of known hosts. Counting objects: 4, done. Delta compression using up to 8 threads. Compressing objects: 100% (3/3), done. Writing objects: 100% (4/4), 2.13 KiB | 0 bytes/s, done. Total 4 (delta 0), reused 0 (delta 0) To git@github.com:DevOpsForVMwareAdministrators/git-test.git * [new branch] master -> master
Figure 3-1 shows what the repository looks like on GitHub after I push the first commit to the remote server.
Figure 3-1 Remote Git repository
If there is a remote repository that you want to use, like the samples we’re providing for this book, you can either use CLI-based methods or GUI-based methods to clone the remote repository to your local machine. Your remote repository system should have options similar to what is pictured on the lower-right corner of Figure 3-1.
The GUI-based methods will differ according to the site you are using. However, on GitHub, you can either use the Clone in Desktop button if you are using its Mac or Windows client, or you can use the Download Zip button, which will be a simple zip file that contains all the source code of the repository. If you are using the Windows or Mac platform, the Clone in Desktop option is recommended because it will create your Git remote link to the remote repository automatically.
The CLI-based method involves taking the SSH or HTTP clone URL and using the git clone command, similar to the following:
git clone https://github.com/DevOpsForVMwareAdministrators/git-test.git
After executing this command, Git will create a directory with the name of the repository (in this case, git-test) and copy the contents of the repository into that directory. You can begin using and updating the code, and the Git remote link will be created for you automatically just like with the Clone in Desktop button mentioned earlier. If you have permission to make changes to the remote repository, you can perform a Git push to forward any local changes to the remote repository. Requesting to make changes to others’ repositories is not within the scope of this book; but if you are interested in the topic, you can research repository forks and pull requests. The implementation of these features may differ between the various public Git repository sites. So, check the appropriate documentation for the site that you are using.
If you are having issues working with the remote repository setup, an alternate workflow consists of the following:
- Before you begin your work, create an empty repository on your remote Git site of choice (for instance, GitHub).
- Clone the empty repository to your local machine using the GUI or CLI methods discussed earlier.
- Begin working in the cloned directory and perform your commits there. You can then use the same git push command introduced earlier to forward your source code commits to the remote repository.
Summary
We discussed the foundational tools that we will use throughout the book to build our test environments, namely Vagrant and Git. Vagrant is useful for building your test environments quickly, and Git will help you keep track of changes that you make to your environment definitions in case you need to undo a change. In the next chapter, we begin discussing the Puppet configuration management technology with hands-on examples.