How to build your inventory
Ansible works against multiple managed nodes or “hosts” in your infrastructure at the same time, using a list or group of lists known as inventory. Once your inventory is defined, you use patterns to select the hosts or groups you want Ansible to run against.
The default location for inventory is a file called
/etc/ansible/hosts. You can specify a different inventory file at the command line using the
-i <path> option. You can also use multiple inventory files at the same time as described in Using multiple inventory sources, and/or pull inventory from dynamic or cloud sources or different formats (YAML, ini, and so on), as described in Working with dynamic inventory.
Introduced in version 2.4, Ansible has Inventory plugins to make this flexible and customizable.
The inventory file can be in one of many formats, depending on the inventory plugins you have.
The most common formats are INI and YAML. A basic INI
/etc/ansible/hosts might look like this:
mail.example.com [webservers] foo.example.com bar.example.com [dbservers] one.example.com two.example.com three.example.com
The headings in brackets are group names, which are used in classifying hosts and deciding what hosts you are controlling at what times and for what purpose. Group names should follow the same guidelines as Creating valid variable names.
Here’s that same basic inventory file in YAML format:
all: hosts: mail.example.com: children: webservers: hosts: foo.example.com: bar.example.com: dbservers: hosts: one.example.com: two.example.com: three.example.com:
There are two default groups:
all group contains every host.
ungrouped group contains all hosts that don’t have another group aside from
Every host will always belong to at least 2 groups (
all and some other group). Though
ungrouped are always present, they can be implicit and not appear in group listings like
You can (and probably will) put each host in more than one group. For example a production webserver in a datacenter in Atlanta might be included in groups called [prod] and [atlanta] and [webservers]. You can create groups that track:
What - An application, stack or microservice (for example, database servers, web servers, and so on).
Where - A datacenter or region, to talk to local DNS, storage, and so on (for example, east, west).
When - The development stage, to avoid testing on production resources (for example, prod, test).
Extending the previous YAML inventory to include what, when, and where would look like:
all: hosts: mail.example.com: children: webservers: hosts: foo.example.com: bar.example.com: dbservers: hosts: one.example.com: two.example.com: three.example.com: east: hosts: foo.example.com: one.example.com: two.example.com: west: hosts: bar.example.com: three.example.com: prod: hosts: foo.example.com: one.example.com: two.example.com: test: hosts: bar.example.com: three.example.com:
You can see that
one.example.com exists in the
You can also use nested groups to simplify
test in this inventory, for the same result:
all: hosts: mail.example.com: children: webservers: hosts: foo.example.com: bar.example.com: dbservers: hosts: one.example.com: two.example.com: three.example.com: east: hosts: foo.example.com: one.example.com: two.example.com: west: hosts: bar.example.com: three.example.com: prod: children: east: test: children: west:
You can find more examples on how to organize your inventories and group your hosts in Inventory setup examples.
If you have a lot of hosts with a similar pattern, you can add them as a range rather than listing each hostname separately:
... webservers: hosts: www[01:50].example.com:
You can specify a stride (increments between sequence numbers) when defining a numeric range of hosts:
... webservers: hosts: www[01:50:2].example.com:
The example above would make the subdomains www01, www03, www05, …, www49 match, but not www00, www02, www50 and so on, because the stride (increment) is 2 units each step.
For numeric patterns, leading zeros can be included or removed, as desired. Ranges are inclusive. You can also define alphabetic ranges:
You can store variable values that relate to a specific host or group in inventory. To start with, you may add variables directly to the hosts and groups in your main inventory file. As you add more and more managed nodes to your Ansible inventory, however, you will likely want to store variables in separate host and group variable files. See Defining variables in inventory for details.
You can easily assign a variable to a single host, then use it later in playbooks. In INI:
[atlanta] host1 http_port=80 maxRequestsPerChild=808 host2 http_port=303 maxRequestsPerChild=909
atlanta: hosts: host1: http_port: 80 maxRequestsPerChild: 808 host2: http_port: 303 maxRequestsPerChild: 909
Unique values like non-standard SSH ports work well as host variables. You can add them to your Ansible inventory by adding the port number after the hostname with a colon:
Connection variables also work well as host variables:
[targets] localhost ansible_connection=local other1.example.com ansible_connection=ssh ansible_user=myuser other2.example.com ansible_connection=ssh ansible_user=myotheruser
If you list non-standard SSH ports in your SSH config file, the
openssh connection will find and use them, but the
paramiko connection will not.
You can also define aliases in your inventory:
jumper ansible_port=5555 ansible_host=192.0.2.50
... hosts: jumper: ansible_port: 5555 ansible_host: 192.0.2.50
In the above example, running Ansible against the host alias “jumper” will connect to 192.0.2.50 on port 5555. See behavioral inventory parameters to further customize the connection to hosts.
Values passed in the INI format using the
key=value syntax are interpreted differently depending on where they are declared:
When declared inline with the host, INI values are interpreted as Python literal structures (strings, numbers, tuples, lists, dicts, booleans, None). Host lines accept multiple
key=valueparameters per line. Therefore they need a way to indicate that a space is part of a value rather than a separator. Values that contain whitespace can be quoted (single or double). See the Python shlex parsing rules for details.
When declared in a
:varssection, INI values are interpreted as strings. For example
var=FALSEwould create a string equal to ‘FALSE’. Unlike host lines,
:varssections accept only a single entry per line, so everything after the
=must be the value for the entry.
If a variable value set in an INI inventory must be a certain type (for example, a string or a boolean value), always specify the type with a filter in your task. Do not rely on types set in INI inventories when consuming variables.
Consider using YAML format for inventory sources to avoid confusion on the actual type of a variable. The YAML inventory plugin processes variable values consistently and correctly.
Generally speaking, this is not the best way to define variables that describe your system policy. Setting variables in the main inventory file is only a shorthand. See Organizing host and group variables for guidelines on storing variable values in individual files in the ‘host_vars’ directory.
If all hosts in a group share a variable value, you can apply that variable to an entire group at once. In INI:
[atlanta] host1 host2 [atlanta:vars] ntp_server=ntp.atlanta.example.com proxy=proxy.atlanta.example.com
atlanta: hosts: host1: host2: vars: ntp_server: ntp.atlanta.example.com proxy: proxy.atlanta.example.com
Group variables are a convenient way to apply variables to multiple hosts at once. Before executing, however, Ansible always flattens variables, including inventory variables, to the host level. If a host is a member of multiple groups, Ansible reads variable values from all of those groups. If you assign different values to the same variable in different groups, Ansible chooses which value to use based on internal rules for merging.
You can make groups of groups using the
:children suffix in INI or the
children: entry in YAML.
You can apply variables to these groups of groups using
[atlanta] host1 host2 [raleigh] host2 host3 [southeast:children] atlanta raleigh [southeast:vars] some_server=foo.southeast.example.com halon_system_timeout=30 self_destruct_countdown=60 escape_pods=2 [usa:children] southeast northeast southwest northwest
all: children: usa: children: southeast: children: atlanta: hosts: host1: host2: raleigh: hosts: host2: host3: vars: some_server: foo.southeast.example.com halon_system_timeout: 30 self_destruct_countdown: 60 escape_pods: 2 northeast: northwest: southwest:
If you need to store lists or hash data, or prefer to keep host and group specific variables separate from the inventory file, see Organizing host and group variables.
Child groups have a couple of properties to note:
Any host that is member of a child group is automatically a member of the parent group.
A child group’s variables will have higher precedence (override) a parent group’s variables.
Groups can have multiple parents and children, but not circular relationships.
Hosts can also be in multiple groups, but there will only be one instance of a host, merging the data from the multiple groups.
Although you can store variables in the main inventory file, storing separate host and group variables files may help you organize your variable values more easily. Host and group variable files must use YAML syntax. Valid file extensions include ‘.yml’, ‘.yaml’, ‘.json’, or no file extension. See YAML Syntax if you are new to YAML.
Ansible loads host and group variable files by searching paths relative to the inventory file or the playbook file. If your inventory file at
/etc/ansible/hosts contains a host named ‘foosball’ that belongs to two groups, ‘raleigh’ and ‘webservers’, that host will use variables in YAML files at the following locations:
/etc/ansible/group_vars/raleigh # can optionally end in '.yml', '.yaml', or '.json' /etc/ansible/group_vars/webservers /etc/ansible/host_vars/foosball
For example, if you group hosts in your inventory by datacenter, and each datacenter uses its own NTP server and database server, you can create a file called
/etc/ansible/group_vars/raleigh to store the variables for the
--- ntp_server: acme.example.org database_server: storage.example.org
You can also create directories named after your groups or hosts. Ansible will read all the files in these directories in lexicographical order. An example with the ‘raleigh’ group:
All hosts in the ‘raleigh’ group will have the variables defined in these files available to them. This can be very useful to keep your variables organized when a single file gets too big, or when you want to use Ansible Vault on some group variables.
ansible-playbook you can also add
host_vars/ directories to your playbook directory. Other Ansible commands (for example,
ansible-console, and so on) will only look for
host_vars/ in the inventory directory. If you want other commands to load group and host variables from a playbook directory, you must provide the
--playbook-dir option on the command line.
If you load inventory files from both the playbook directory and the inventory directory, variables in the playbook directory will override variables set in the inventory directory.
Keeping your inventory file and variables in a git repo (or other version control) is an excellent way to track changes to your inventory and host variables.
By default variables are merged/flattened to the specific host before a play is run. This keeps Ansible focused on the Host and Task, so groups don’t really survive outside of inventory and host matching. By default, Ansible overwrites variables including the ones defined for a group and/or host (see DEFAULT_HASH_BEHAVIOUR). The order/precedence is (from lowest to highest):
all group (because it is the ‘parent’ of all other groups)
By default Ansible merges groups at the same parent/child level in ASCII order, and the last group loaded overwrites the previous groups. For example, an a_group will be merged with b_group and b_group vars that match will overwrite the ones in a_group.
You can change this behavior by setting the group variable
ansible_group_priority to change the merge order for groups of the same level (after the parent/child order is resolved). The larger the number, the later it will be merged, giving it higher priority. This variable defaults to
1 if not set. For example:
a_group: vars: testvar: a ansible_group_priority: 10 b_group: vars: testvar: b
In this example, if both groups have the same priority, the result would normally have been
testvar == b, but since we are giving the
a_group a higher priority the result will be
testvar == a.
ansible_group_priority can only be set in the inventory source and not in group_vars/, as the variable is used in the loading of group_vars.
You can target multiple inventory sources (directories, dynamic inventory scripts
or files supported by inventory plugins) at the same time by giving multiple inventory parameters from the command
line or by configuring
ANSIBLE_INVENTORY. This can be useful when you want to target normally
separate environments, like staging and production, at the same time for a specific action.
Target two sources from the command line like this:
ansible-playbook get_logs.yml -i staging -i production
Keep in mind that if there are variable conflicts in the inventories, they are resolved according
to the rules described in How variables are merged and Variable precedence: Where should I put a variable?.
The merging order is controlled by the order of the inventory source parameters.
[all:vars] in staging inventory defines
myvar = 1, but production inventory defines
myvar = 2,
the playbook will be run with
myvar = 2. The result would be reversed if the playbook was run with
-i production -i staging.
Aggregating inventory sources with a directory
You can also create an inventory by combining multiple inventory sources and source types under a directory. This can be useful for combining static and dynamic hosts and managing them as one inventory. The following inventory combines an inventory plugin source, a dynamic inventory script, and a file with static hosts:
inventory/ openstack.yml # configure inventory plugin to get hosts from Openstack cloud dynamic-inventory.py # add additional hosts with dynamic inventory script static-inventory # add static hosts and groups group_vars/ all.yml # assign variables to all hosts
You can target this inventory directory simply like this:
ansible-playbook example.yml -i inventory
It can be useful to control the merging order of the inventory sources if there’s variable conflicts or group of groups dependencies to the other inventory sources. The inventories are merged in ASCII order according to the filenames so the result can be controlled by adding prefixes to the files:
inventory/ 01-openstack.yml # configure inventory plugin to get hosts from Openstack cloud 02-dynamic-inventory.py # add additional hosts with dynamic inventory script 03-static-inventory # add static hosts group_vars/ all.yml # assign variables to all hosts
myvar = 1 for the group
myvar = 2,
myvar = 3, the playbook will be run with
myvar = 3.
As described above, setting the following variables control how Ansible interacts with remote hosts.
Ansible does not expose a channel to allow communication between the user and the ssh process to accept a password manually to decrypt an ssh key when using the ssh connection plugin (which is the default). The use of
ssh-agent is highly recommended.
Connection type to the host. This can be the name of any of ansible’s connection plugins. SSH protocol types are
paramiko. The default is smart. Non-SSH based types are described in the next section.
General for all connections:
The name of the host to connect to, if different from the alias you wish to give to it.
The connection port number, if not the default (22 for ssh)
The user name to use when connecting to the host
The password to use to authenticate to the host (never store this variable in plain text; always use a vault. See Keep vaulted variables safely visible)
Specific to the SSH connection:
Private key file used by ssh. Useful if using multiple keys and you don’t want to use SSH agent.
This setting is always appended to the default command line for sftp, scp, and ssh. Useful to configure a
ProxyCommandfor a certain host (or group).
This setting is always appended to the default sftp command line.
This setting is always appended to the default scp command line.
This setting is always appended to the default ssh command line.
Determines whether or not to use SSH pipelining. This can override the
- ansible_ssh_executable (added in version 2.2)
This setting overrides the default behavior to use the system ssh. This can override the
Privilege escalation (see Ansible Privilege Escalation for further details):
ansible_su, allows to force privilege escalation
Allows to set privilege escalation method
ansible_su_user, allows to set the user you become through privilege escalation
ansible_su_password, allows you to set the privilege escalation password (never store this variable in plain text; always use a vault. See Keep vaulted variables safely visible)
ansible_su_exe, allows you to set the executable for the escalation method selected
ansible_su_flags, allows you to set the flags passed to the selected escalation method. This can be also set globally in
Remote host environment parameters:
The shell type of the target system. You should not use this setting unless you have set the ansible_shell_executable to a non-Bourne (sh) compatible shell. By default commands are formatted using
sh-style syntax. Setting this to
fishwill cause commands executed on target systems to follow those shell’s syntax instead.
The target host python path. This is useful for systems with more than one Python or not located at /usr/bin/python such as *BSD, or where /usr/bin/python is not a 2.X series Python. We do not use the /usr/bin/env mechanism as that requires the remote user’s path to be set right and also assumes the python executable is named python, where the executable might be named something like python2.6.
Works for anything such as ruby or perl and works just like ansible_python_interpreter. This replaces shebang of modules which will run on that host.
New in version 2.1.
This sets the shell the ansible controller will use on the target machine, overrides
ansible.cfgwhich defaults to /bin/sh. You should really only change it if is not possible to use /bin/sh (in other words, if /bin/sh is not installed on the target machine or cannot be run from sudo.).
Examples from an Ansible-INI host file:
some_host ansible_port=2222 ansible_user=manager aws_host ansible_ssh_private_key_file=/home/example/.ssh/aws.pem freebsd_host ansible_python_interpreter=/usr/local/bin/python ruby_module_host ansible_ruby_interpreter=/usr/bin/ruby.1.9.3
As stated in the previous section, Ansible executes playbooks over SSH but it is not limited to this connection type.
With the host specific parameter
ansible_connection=<connector>, the connection type can be changed.
The following non-SSH based connectors are available:
This connector can be used to deploy the playbook to the control machine itself.
This connector deploys the playbook directly into Docker containers using the local Docker client. The following parameters are processed by this connector:
The name of the Docker container to connect to.
The user name to operate within the container. The user must exist inside the container.
If set to
become_userwill be used to operate within the container.
Could be a string with any additional arguments understood by Docker, which are not command specific. This parameter is mainly used to configure a remote Docker daemon to use.
Here is an example of how to instantly deploy to created containers:
- name: Create a jenkins container community.general.docker_container: docker_host: myserver.net:4243 name: my_jenkins image: jenkins - name: Add the container to inventory ansible.builtin.add_host: name: my_jenkins ansible_connection: docker ansible_docker_extra_args: "--tlsverify --tlscacert=/path/to/ca.pem --tlscert=/path/to/client-cert.pem --tlskey=/path/to/client-key.pem -H=tcp://myserver.net:4243" ansible_user: jenkins changed_when: false - name: Create a directory for ssh keys delegate_to: my_jenkins ansible.builtin.file: path: "/var/jenkins_home/.ssh/jupiter" state: directory
For a full list with available plugins and examples, see Plugin list.
If you’re reading the docs from the beginning, this may be the first example you’ve seen of an Ansible playbook. This is not an inventory file. Playbooks will be covered in great detail later in the docs.
See also Sample Ansible setup, which shows inventory along with playbooks and other Ansible artifacts.
If you need to manage multiple environments it’s sometimes prudent to have only hosts of a single environment defined per inventory. This way, it is harder to, for instance, accidentally change the state of nodes inside the “test” environment when you actually wanted to update some “staging” servers.
For the example mentioned above you could have an
[dbservers] db01.test.example.com db02.test.example.com [appservers] app01.test.example.com app02.test.example.com app03.test.example.com
That file only includes hosts that are part of the “test”
environment. Define the “staging” machines in another file
[dbservers] db01.staging.example.com db02.staging.example.com [appservers] app01.staging.example.com app02.staging.example.com app03.staging.example.com
To apply a playbook called
to all the app servers in the test environment, use the
ansible-playbook -i inventory_test -l appservers site.yml
In the previous section you already saw an example for using groups in order to cluster hosts that have the same function. This allows you, for instance, to define firewall rules inside a playbook or role affecting only database servers:
- hosts: dbservers tasks: - name: Allow access from 10.0.0.1 ansible.builtin.iptables: chain: INPUT jump: ACCEPT source: 10.0.0.1
Other tasks might be focused on where a certain host is located. Let’s
located in DC1 while
db02.test.example.com is in DC2:
[dc1] db01.test.example.com app01.test.example.com [dc2] db02.test.example.com
In practice, you might even end up mixing all these setups as you might need to, on one day, update all nodes in a specific data center while, on another day, update all the application servers no matter their location.
- Inventory plugins
Pulling inventory from dynamic or static sources
- Working with dynamic inventory
Pulling inventory from dynamic sources, such as cloud providers
- Introduction to ad hoc commands
Examples of basic commands
- Working with playbooks
Learning Ansible’s configuration, deployment, and orchestration language.
- Mailing List
Questions? Help? Ideas? Stop by the list on Google Groups
- Real-time chat
How to join Ansible chat channels