Update Ansible provisioner to only create a single inventory file

The Ansible provisioner will now only create a single inventory file named,
"vagrant_ansible_inventory". All defined Vagrant machines will be added to
this inventory file. Provisioning will now include a "--limit=#{machine}"
option to scope Ansible provisioning tasks to just the current machine. Setting
the Ansible provisioner's "limit" config option will override the new default
limit. Ansible provisioning scripts will now have access to all other defined
machines and what groups they reside in.

plugins/provisioners/ansible/provisioner.rb

@@ -13,6 +13,14 @@ module VagrantPlugins
         # Joker! Not (yet) supported arguments can be passed this way.
         options.concat(self.as_array(config.raw_arguments)) if config.raw_arguments
 
+        # By default we limit by the current machine.  This can be
+        # overriden by the limit config option.
+        limit_option = if config.limit == nil
+                         "--limit=#{@machine.name}"
+                       elsif not config.limit.empty?
+                         "--limit=#{as_list_argument(config.limit)}"
+                       end
+
         # Append Provisioner options (highest precedence):
         options << "--inventory-file=#{self.setup_inventory_file}"
         options << "--extra-vars=#{self.get_extra_vars_argument}" if config.extra_vars
@@ -22,7 +30,7 @@ module VagrantPlugins
         options << "--ask-sudo-pass" if config.ask_sudo_pass
         options << "--tags=#{as_list_argument(config.tags)}" if config.tags
         options << "--skip-tags=#{as_list_argument(config.skip_tags)}" if config.skip_tags
-        options << "--limit=#{as_list_argument(config.limit)}" if config.limit
+        options << limit_option if limit_option
         options << "--start-at-task=#{config.start_at_task}" if config.start_at_task
 
         # Assemble the full ansible-playbook command
@@ -64,25 +72,31 @@ module VagrantPlugins
         ssh = @machine.ssh_info
 
         generated_inventory_file =
-          @machine.env.root_path.join("vagrant_ansible_inventory_#{machine.name}")
+          @machine.env.root_path.join("vagrant_ansible_inventory")
 
         generated_inventory_file.open('w') do |file|
           file.write("# Generated by Vagrant\n\n")
-          file.write("#{machine.name} ansible_ssh_host=#{ssh[:host]} ansible_ssh_port=#{ssh[:port]}\n")
 
-          # Write out groups information.  Only include current
+          @machine.env.active_machines.each do |am|
-          # machine and its groups to avoid Ansible errors on
+            m = @machine.env.machine(*am)
-          # provisioning.
+            file.write("#{m.name} ansible_ssh_host=#{m.ssh_info[:host]} ansible_ssh_port=#{m.ssh_info[:port]}\n")
+          end
+
+          # Write out groups information.  Only includes groups and
+          # machines which have been defined, otherwise Ansible will
+          # complain.
           groups_of_groups = {}
           included_groups = []
 
           config.groups.each_pair do |gname, gmembers|
             if gname.end_with?(":children")
               groups_of_groups[gname] = gmembers
-            elsif gmembers.include?("#{machine.name}")
+            else
               included_groups << gname
               file.write("\n[#{gname}]\n")
-              file.write("#{machine.name}\n")
+              gmembers.each do |gm|
+                file.write("#{gm}\n") if @machine.env.active_machines.map { |m| m[0] }.include?(gm.to_sym)
+              end
             end
           end
 

website/docs/source/v2/provisioning/ansible.html.md

@@ -30,15 +30,15 @@ a single page of documentation.
 When using Ansible, it needs to know on which machines a given playbook should run. It does
 this by way of an inventory file which lists those machines. In the context of Vagrant,
 there are two ways to approach working with inventory files. The first and simplest option
-is to not provide one to Vagrant at all. Vagrant will generate inventory files for each
+is to not provide one to Vagrant at all. Vagrant will generate an inventory file encompassing
-virtual machine it manages, and use them for provisioning machines. Generated inventory files
+all of the virtual machine it manages, and use it for provisioning machines. The generated
-are created adjacent to your Vagrantfile, named using the machine names set in your Vagrantfile.
+inventory file is created adjacent to your Vagrantfile, named `vagrant_ansible_inventory`.
 
-The second option is for situations where you'd like to have more than one virtual machine
+The second option is for situations where you'd like to have more control over the inventory file,
-in a single inventory file, perhaps leveraging more complex playbooks or inventory grouping.
+perhaps leveraging more complex playbooks or inventory grouping. If you provide the
-If you provide the `ansible.inventory_path` option referencing a specific inventory file
+`ansible.inventory_path` option referencing a specific inventory file dedicated to your Vagrant
-dedicated to your Vagrant project, that one will be used instead of generating them.
+project, that one will be used instead of generating them. Such an inventory file for use with
-Such an inventory file for use with Vagrant might look like:
+Vagrant might look like:
 
 ```
 [vagrant]
@@ -153,8 +153,8 @@ by the sudo command.
     ansible.groups = {
       "group1" => ["machine1"],
       "group2" => ["machine2", "machine3"],
-      "all_groups:children" => ["group1", "group2"]
+      "all_groups:children" => ["group1", "group2", "group3"]
     }
     ```
-    Note that only the current machine and its related groups will be added to the inventory file.
+    Note that undefined machines and groups are not added to the inventory.  For example, `group3` in the above example would not be added to the inventory file.
 * `ansible.host_key_checking` can be set to `false` which will disable host key checking and prevent `"FAILED: (25, 'Inappropriate ioctl for device')"` errors from being reported during provisioner runs.  The default value is `true`, which matches the default behavior of Ansible 1.2.1 and later.