5.7 KiB

Raw Blame History

page_title	sidebar_current
Basic Usage - Provisioning	provisioning-basic

Basic Usage of Provisioners

While Vagrant offers multiple options for how you are able to provision your machine, there is a standard usage pattern as well as some important points common to all provisioners that are important to know.

Configuration

First, every provisioner is configured within your Vagrantfile using the config.vm.provision method call. For example, the Vagrantfile below enables shell provisioning:

Vagrant.configure("2") do |config|
  # ... other configuration

  config.vm.provision "shell", inline: "echo hello"
end

Every provisioner has a type, such as "shell", used as the first parameter to the provisioning configuration. Following that is basic key/value for configuring that specific provisioner. Instead of basic key/value, you can also use a Ruby block for a syntax that is more like variable assignment. The following is effectively the same as the prior example:

Vagrant.configure("2") do |config|
  # ... other configuration

  config.vm.provision "shell" do |s|
    s.inline = "echo hello"
  end
end

The benefit of the block-based syntax is that with more than a couple options it can greatly improve readability. Additionally, some provisioners, like the Chef provisioner, have special methods that can be called within that block to ease configuration that can't be done with the key/value approach.

Running Provisioners

Provisioners are run in three cases: the initial vagrant up, vagrant provision, and vagrant reload --provision.

A --no-provision flag can be passed to up and reload if you don't want to run provisioners. Likewise, you can pass --provision to force provisioning.

The --provision-with flag can be used if you only want to run a specific provisioner if you have multiple provisioners specified. For example, if you have a shell and Puppet provisioner and only want to run the shell one, you can do vagrant provision --provision-with shell.

Run Once or Always

By default, provisioners are only run once, during the first vagrant up since the last vagrant destroy, unless the --provision flag is set, as noted above.

Optionally, you can configure provisioners to run on every up or reload. They'll only be not run if the --no-provision flag is explicitly specified. To do this set the run option to "always", as shown below:

Vagrant.configure("2") do |config|
  config.vm.provision "shell", inline: "echo hello",
    run: "always"
end

If you're using the block format, you must specify it outside of the block, as shown below:

Vagrant.configure("2") do |config|
  config.vm.provision "shell", run: "always" do |s|
    s.inline = "echo hello"
  end
end

Multiple Provisioners

Multiple config.vm.provision methods can be used to define multiple provisioners. These provisioners will be run in the order they're defined. This is useful for a variety of reasons, but most commonly it is used so that a shell script can bootstrap some of the system so that another provisioner can take over later.

If you define provisioners at multiple "scope" levels (such as globally in the configuration block, then in a multi-machine definition, then maybe in a provider-specific override), then the outer scopes will always run before any inner scopes. For example, in the Vagrantfile below:

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

  config.vm.define "web" do |web|
    web.vm.provision "shell", inline: "echo bar"
  end

  config.vm.provision "shell", inline: "echo baz"
end

The ordering of the provisioners will be to echo "foo", "baz", then "bar" (note the second one might not be what you expect!). Remember: ordering is outside in.

Overriding Provisioner Settings

Warning: Advanced Topic! Provisioner overriding is an advanced topic that really only becomes useful if you're already using multi-machine and/or provider overrides. If you're just getting started with Vagrant, you can safely skip this.

When using features such as multi-machine or provider-specific overrides, you may want to define common provisioners in the global configuration scope of a Vagrantfile, but override certain aspects of them internally. Vagrant allows you to do this, but has some details to consider.

To override settings, you must assign an ID to your provisioner. Then it is only a matter of specifying the same ID to override:

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

  config.vm.define "web" do |web|
    web.vm.provision "shell",
      inline: "echo bar", id: "foo"
  end
end

In the above, only "bar" will be echoed, because the inline setting overloaded the outer provisioner. This overload is only effective within that scope: the "web" VM. If there were another VM defined, it would still echo "foo" unless it itself also overloaded the provisioner.

Be careful with ordering. When overriding a provisioner in a sub-scope, the provisioner will run at that point. In the example below, the output would be "foo" then "bar":

Vagrant.configure("2") do |config|
  config.vm.provision "shell",
    inline: "echo ORIGINAL!", id: "foo"

  config.vm.define "web" do |web|
    web.vm.provision "shell",
      inline: "echo foo"
    web.vm.provision "shell",
      inline: "echo bar", id: "foo"
  end
end

If you want to preserve the original ordering, you can specify the preserve_order: true flag.

5.7 KiB Raw Blame History