Multi-Machine

Vagrant is able to define and control multiple guest machines per Vagrantfile. This is known as a "multi-machine" environment.

These machines are generally able to work together or are somehow associated with each other. Here are some use-cases people are using multi-machine environments for today:

  • Accurately modeling a multi-server production topology, such as separating a web and database server.
  • Modeling a distributed system and how they interact with each other.
  • Testing an interface, such as an API to a service component.
  • Disaster-case testing: machines dying, network partitions, slow networks, inconsistent world views, etc.

Historically, running complex environments such as these was done by flattening them onto a single machine. The problem with that is that it is an inaccurate model of the production setup, which can behave far differently.

Using the multi-machine feature of Vagrant, these environments can be modeled in the context of a single Vagrant environment without losing any of the benefits of Vagrant.

Defining Multiple Machines

Multiple machines are defined within the same project Vagrantfile using the config.vm.define method call. This configuration directive is a little funny, because it creates a Vagrant configuration within a configuration. An example shows this best:

Vagrant.configure("2") do |config|
  config.vm.provision "shell", inline: "echo Hello"

  config.vm.define "web" do |web|
    web.vm.box = "apache"
  end

  config.vm.define "db" do |db|
    db.vm.box = "mysql"
  end
end

As you can see, config.vm.define takes a block with another variable. This variable, such as web above, is the exact same as the config variable, except any configuration of the inner variable applies only to the machine being defined. Therefore, any configuration on web will only affect the web machine.

And importantly, you can continue to use the config object as well. The configuration object is loaded and merged before the machine-specific configuration, just like other Vagrantfiles within the Vagrantfile load order.

If you're familiar with programming, this is similar to how languages have different variable scopes.

When using these scopes, order of execution for things such as provisioners becomes important. Vagrant enforces ordering outside-in, in the order listed in the Vagrantfile. For example, with the Vagrantfile below:

Vagrant.configure("2") do |config|
    config.vm.provision :shell, inline: 'echo A'
    config.vm.define :testing do |test|
        test.vm.provision :shell, inline: 'echo B'
    end
    config.vm.provision :shell, inline: 'echo C'
end

The provisioners in this case will output "A", then "C", then "B". Notice that "B" is last. That is because the ordering is outside-in, in the order of the file.

Controlling Multiple Machines

The moment more than one machine is defined within a Vagrantfile, the usage of the various vagrant commands changes slightly. The change should be mostly intuitive.

Commands that only make sense to target a single machine, such as vagrant ssh, now require the name of the machine to control. Using the example above, you would say vagrant ssh web or vagrant ssh db.

Other commands, such as vagrant up, operate on every machine by default. So if you ran vagrant up, Vagrant would bring up both the web and DB machine. You could also optionally be specific and say vagrant up web or vagrant up db.

Additionally, you can specify a regular expression for matching only certain machines. This is useful in some cases where you specify many similar machines, for example if you're testing a distributed service you may have a leader machine as well as a follower0, follower1, follower2, etc. If you want to bring up all the followers but not the leader, you can just do vagrant up /follower[0-9]/. If Vagrant sees a machine name within forward slashes, it assumes you're using a regular expression.

Communication Between Machines

In order to facilitate communication within machines in a multi-machine setup, the various networking options should be used. In particular, the private network can be used to make a private network between multiple machines and the host.

Specifying a Primary Machine

You can also specify a primary machine. The primary machine will be the default machine used when a specific machine in a multi-machine environment is not specified.

To specify a default machine, just mark it primary when defining it. Only one primary machine may be specified.

config.vm.define "web", primary: true do |web|
  # ...
end

Autostart Machines

By default in a multi-machine environment, vagrant up will start all of the defined machines. The autostart setting allows you to tell Vagrant to not start specific machines. Example:

config.vm.define "web"
config.vm.define "db"
config.vm.define "db_follower", autostart: false

When running vagrant up with the settings above, Vagrant will automatically start the "web" and "db" machines, but will not start the "db_follower" machine. You can manually force the "db_follower" machine to start by running vagrant up db_follower.