Ansible 2.5 Porting Guide
This section discusses the behavioral changes between Ansible 2.4 and Ansible 2.5.
It is intended to assist in updating your playbooks, plugins and other parts of your Ansible infrastructure so they will work with this version of Ansible.
We suggest you read this page along with Ansible Changelog for 2.5 to understand what updates you may need to make.
This document is part of a collection on porting. The complete list of porting guides can be found at porting guides.
Playbook
Dynamic includes and attribute inheritance
In Ansible version 2.4, the concept of dynamic includes (include_tasks), as opposed to static imports (import_tasks), was introduced to clearly define the differences in how include works between dynamic and static includes.
All attributes applied to a dynamic include_* would only apply to the include itself, while attributes applied to a
static import_* would be inherited by the tasks within.
This separation was only partially implemented in Ansible version 2.4. As of Ansible version 2.5, this work is complete and the separation now behaves as designed; attributes applied to an include_* task will not be inherited by the tasks within.
To achieve an outcome similar to how Ansible worked prior to version 2.5, playbooks should use an explicit application of the attribute on the needed tasks, or use blocks to apply the attribute to many tasks. Another option is to use a static import_* when possible instead of a dynamic task.
OLD In Ansible 2.4:
- include_tasks: "{{ ansible_distribution }}.yml"
tags:
- distro_include
Included file:
- block:
- debug:
msg: "In included file"
- apt:
name: nginx
state: latest
NEW In Ansible 2.5:
Including task:
- include_tasks: "{{ ansible_distribution }}.yml"
tags:
- distro_include
Included file:
- block:
- debug:
msg: "In included file"
- apt:
name: nginx
state: latest
tags:
- distro_include
The relevant change in those examples is, that in Ansible 2.5, the included file defines the tag distro_include again. The tag is not inherited automatically.
Fixed handling of keywords and inline variables
We made several fixes to how we handle keywords and ‘inline variables’, to avoid conflating the two. Unfortunately these changes mean you must specify whether name is a keyword or a variable when calling roles. If you have playbooks that look like this:
roles:
- { role: myrole, name: Justin, othervar: othervalue, become: True}
You will run into errors because Ansible reads name in this context as a keyword. Beginning in 2.5, if you want to use a variable name that is also a keyword, you must explicitly declare it as a variable for the role:
roles:
- { role: myrole, vars: {name: Justin, othervar: othervalue}, become: True}
For a full list of keywords see Playbook Keywords.
Migrating from with_X to loop
In most cases, loops work best with the loop keyword instead of with_X style loops. The loop syntax is usually best expressed using filters instead of more complex use of query or lookup.
These examples show how to convert many common with_ style loops to loop and filters.
with_list
with_list is directly replaced by loop.
- name: with_list
ansible.builtin.debug:
msg: "{{ item }}"
with_list:
- one
- two
- name: with_list -> loop
ansible.builtin.debug:
msg: "{{ item }}"
loop:
- one
- two
with_items
with_items is replaced by loop and the flatten filter.
- name: with_items
ansible.builtin.debug:
msg: "{{ item }}"
with_items: "{{ items }}"
- name: with_items -> loop
ansible.builtin.debug:
msg: "{{ item }}"
loop: "{{ items|flatten(levels=1) }}"
with_indexed_items
with_indexed_items is replaced by loop, the flatten filter and loop_control.index_var.
- name: with_indexed_items
ansible.builtin.debug:
msg: "{{ item.0 }} - {{ item.1 }}"
with_indexed_items: "{{ items }}"
- name: with_indexed_items -> loop
ansible.builtin.debug:
msg: "{{ index }} - {{ item }}"
loop: "{{ items|flatten(levels=1) }}"
loop_control:
index_var: index
with_flattened
with_flattened is replaced by loop and the flatten filter.
- name: with_flattened
ansible.builtin.debug:
msg: "{{ item }}"
with_flattened: "{{ items }}"
- name: with_flattened -> loop
ansible.builtin.debug:
msg: "{{ item }}"
loop: "{{ items|flatten }}"
with_together
with_together is replaced by loop and the zip filter.
- name: with_together
ansible.builtin.debug:
msg: "{{ item.0 }} - {{ item.1 }}"
with_together:
- "{{ list_one }}"
- "{{ list_two }}"
- name: with_together -> loop
ansible.builtin.debug:
msg: "{{ item.0 }} - {{ item.1 }}"
loop: "{{ list_one|zip(list_two)|list }}"
Another example with complex data
- name: with_together -> loop
ansible.builtin.debug:
msg: "{{ item.0 }} - {{ item.1 }} - {{ item.2 }}"
loop: "{{ data[0]|zip(*data[1:])|list }}"
vars:
data:
- ['a', 'b', 'c']
- ['d', 'e', 'f']
- ['g', 'h', 'i']
with_dict
with_dict can be substituted by loop and either the dictsort or dict2items filters.
- name: with_dict
ansible.builtin.debug:
msg: "{{ item.key }} - {{ item.value }}"
with_dict: "{{ dictionary }}"
- name: with_dict -> loop (option 1)
ansible.builtin.debug:
msg: "{{ item.key }} - {{ item.value }}"
loop: "{{ dictionary|dict2items }}"
- name: with_dict -> loop (option 2)
ansible.builtin.debug:
msg: "{{ item.0 }} - {{ item.1 }}"
loop: "{{ dictionary|dictsort }}"
with_sequence
with_sequence is replaced by loop and the range function, and potentially the format filter.
- name: with_sequence
ansible.builtin.debug:
msg: "{{ item }}"
with_sequence: start=0 end=4 stride=2 format=testuser%02x
- name: with_sequence -> loop
ansible.builtin.debug:
msg: "{{ 'testuser%02x' | format(item) }}"
loop: "{{ range(0, 4 + 1, 2)|list }}"
The range of the loop is exclusive of the end point.
with_subelements
with_subelements is replaced by loop and the subelements filter.
- name: with_subelements
ansible.builtin.debug:
msg: "{{ item.0.name }} - {{ item.1 }}"
with_subelements:
- "{{ users }}"
- mysql.hosts
- name: with_subelements -> loop
ansible.builtin.debug:
msg: "{{ item.0.name }} - {{ item.1 }}"
loop: "{{ users|subelements('mysql.hosts') }}"
with_nested/with_cartesian
with_nested and with_cartesian are replaced by loop and the product filter.
- name: with_nested
ansible.builtin.debug:
msg: "{{ item.0 }} - {{ item.1 }}"
with_nested:
- "{{ list_one }}"
- "{{ list_two }}"
- name: with_nested -> loop
ansible.builtin.debug:
msg: "{{ item.0 }} - {{ item.1 }}"
loop: "{{ list_one|product(list_two)|list }}"
with_random_choice
with_random_choice is replaced by just use of the random filter, without need of loop.
- name: with_random_choice
ansible.builtin.debug:
msg: "{{ item }}"
with_random_choice: "{{ my_list }}"
- name: with_random_choice -> loop (No loop is needed here)
ansible.builtin.debug:
msg: "{{ my_list|random }}"
tags: random
Deprecated
Jinja tests used as filters
Using Ansible-provided jinja tests as filters will be removed in Ansible 2.9.
Prior to Ansible 2.5, jinja tests included within Ansible were most often used as filters. The large difference in use is that filters are referenced as variable | filter_name while jinja tests are referenced as variable is test_name.
Jinja tests are used for comparisons, while filters are used for data manipulation and have different applications in jinja. This change is to help differentiate the concepts for a better understanding of jinja, and where each can be appropriately used.
As of Ansible 2.5, using an Ansible provided jinja test with filter syntax, will display a deprecation error.
OLD In Ansible 2.4 (and earlier) the use of an Ansible included jinja test would likely look like this:
when:
- result | failed
- not result | success
NEW In Ansible 2.5 it should be changed to look like this:
when:
- result is failed
- results is not successful
In addition to the deprecation warnings, many new tests have been introduced that are aliases of the old tests. These new tests make more sense grammatically with the jinja test syntax, such as the new successful test which aliases success.
when: result is successful
See Tests for more information.
Additionally, a script was created to assist in the conversion for tests using filter syntax to proper jinja test syntax. This script has been used to convert all of the Ansible integration tests to the correct format. There are a few limitations documented, and all changes made by this script should be evaluated for correctness before executing the modified playbooks. The script can be found at https://github.com/ansible/ansible/blob/devel/hacking/fix_test_syntax.py.
Ansible fact namespacing
Ansible facts, which have historically been written to names like ansible_*
in the main facts namespace, have been placed in their own new namespace,
ansible_facts.* For example, the fact ansible_distribution is now best
queried through the variable structure ansible_facts.distribution.
A new configuration variable, inject_facts_as_vars, has been added to
ansible.cfg. Its default setting, ‘True’, keeps the 2.4 behavior of facts
variables being set in the old ansible_* locations (while also writing them
to the new namespace). This variable is expected to be set to ‘False’ in a
future release. When inject_facts_as_vars is set to False, you must
refer to ansible_facts through the new ansible_facts.* namespace.
Modules
Major changes in popular modules are detailed here.
github_release
In Ansible versions 2.4 and older, after creating a GitHub release using the create_release state, the github_release module reported state as skipped.
In Ansible version 2.5 and later, after creating a GitHub release using the create_release state, the github_release module now reports state as changed.
Modules removed
The following modules no longer exist:
nxos_mtu use nxos_system’s
system_mtuoption or nxos_interface insteadcl_interface_policy use nclu instead
cl_bridge use nclu instead
cl_img_install use nclu instead
cl_ports use nclu instead
cl_license use nclu instead
cl_interface use nclu instead
cl_bond use nclu instead
ec2_vpc use ec2_vpc_net along with supporting modules ec2_vpc_igw, ec2_vpc_route_table, ec2_vpc_subnet, ec2_vpc_dhcp_option, ec2_vpc_nat_gateway, ec2_vpc_nacl instead.
ec2_ami_search use ec2_ami_facts instead
docker use docker_container and docker_image instead
Note
These modules may no longer have documentation in the current release. Please see the Ansible 2.4 module documentation if you need to know how they worked for porting your playbooks.
Deprecation notices
The following modules will be removed in Ansible 2.9. Please update your playbooks accordingly.
Apstra’s
aos_*modules are deprecated as they do not work with AOS 2.1 or higher. See new modules at https://github.com/apstra.nxos_ip_interface use nxos_l3_interface instead.
nxos_portchannel use nxos_linkagg instead.
nxos_switchport use nxos_l2_interface instead.
panos_security_policy use panos_security_rule instead.
panos_nat_policy use panos_nat_rule instead.
vsphere_guest use vmware_guest instead.
Noteworthy module changes
The stat and win_stat modules have changed the default of the option
get_md5fromtruetofalse.
This option will be removed starting with Ansible version 2.9. The options get_checksum: True
and checksum_algorithm: md5 can still be used if an MD5 checksum is
desired.
osx_saymodule was renamed into say.Several modules which could deal with symlinks had the default value of their
followoption changed as part of a feature to standardize the behavior of follow:The file module changed from
follow=Falsetofollow=Truebecause its purpose is to modify the attributes of a file and most systems do not allow attributes to be applied to symlinks, only to real files.The replace module had its
followparameter removed because it inherently modifies the content of an existing file so it makes no sense to operate on the link itself.The blockinfile module had its
followparameter removed because it inherently modifies the content of an existing file so it makes no sense to operate on the link itself.In Ansible-2.5.3, the template module became more strict about its
srcfile being proper utf-8. Previously, non-utf8 contents in a template module src file would result in a mangled output file (the non-utf8 characters would be replaced with a unicode replacement character). Now, on Python2, the module will error out with the message, “Template source files must be utf-8 encoded”. On Python3, the module will first attempt to pass the non-utf8 characters through verbatim and fail if that does not succeed.
Plugins
As a developer, you can now use ‘doc fragments’ for common configuration options on plugin types that support the new plugin configuration system.
Inventory
Inventory plugins have been fine tuned, and we have started to add some common features:
The ability to use a cache plugin to avoid costly API/DB queries is disabled by default. If using inventory scripts, some may already support a cache, but it is outside of Ansible’s knowledge and control. Moving to the internal cache will allow you to use Ansible’s existing cache refresh/invalidation mechanisms.
A new ‘auto’ plugin, enabled by default, that can automatically detect the correct plugin to use IF that plugin is using our ‘common YAML configuration format’. The previous host_list, script, yaml and ini plugins still work as they did, the auto plugin is now the last one we attempt to use. If you had customized the enabled plugins you should revise the setting to include the new auto plugin.
Shell
Shell plugins have been migrated to the new plugin configuration framework. It is now possible to customize more settings, and settings which were previously ‘global’ can now also be overridden using host specific variables.
For example, system_temps is a new setting that allows you to control what Ansible will consider a ‘system temporary dir’. This is used when escalating privileges for a non-administrative user. Previously this was hardcoded to ‘/tmp’, which some systems cannot use for privilege escalation. This setting now defaults to [ '/var/tmp', '/tmp'].
Another new setting is admin_users which allows you to specify a list of users to be considered ‘administrators’. Previously this was hardcoded to root. It now it defaults to [root, toor, admin]. This information is used when choosing between your remote_temp and system_temps directory.
For a full list, check the shell plugin you are using, the default shell plugin is sh.
Those that had to work around the global configuration limitations can now migrate to a per host/group settings, but also note that the new defaults might conflict with existing usage if the assumptions don’t correlate to your environment.
Filter
The lookup plugin API now throws an error if a non-iterable value is returned from a plugin. Previously, numbers or other non-iterable types returned by a plugin were accepted without error or warning. This change was made because plugins should always return a list. Please note that plugins that return strings and other non-list iterable values will not throw an error, but may cause unpredictable behavior. If you have a custom lookup plugin that does not return a list, you should modify it to wrap the return values in a list.
Lookup
A new option was added to lookup plugins globally named error which allows you to control how errors produced by the lookup are handled, before this option they were always fatal. Valid values for this option are warn, ignore and strict. See the lookup page for more details.
Porting custom scripts
No notable changes.
Network
Expanding documentation
We’re expanding the network documentation. There’s new content and a new Ansible Network landing page. We will continue to build the network-related documentation moving forward.
Top-level connection arguments will be removed in 2.9
Top-level connection arguments like username, host, and password are deprecated and will be removed in version 2.9.
OLD In Ansible < 2.4
- name: example of using top-level options for connection properties
ios_command:
commands: show version
host: "{{ inventory_hostname }}"
username: cisco
password: cisco
authorize: yes
auth_pass: cisco
The deprecation warnings reflect this schedule. The task above, run in Ansible 2.5, will result in:
[DEPRECATION WARNING]: Param 'username' is deprecated. See the module docs for more information. This feature will be removed in version
2.9. Deprecation warnings can be disabled by setting deprecation_warnings=False in ansible.cfg.
[DEPRECATION WARNING]: Param 'password' is deprecated. See the module docs for more information. This feature will be removed in version
2.9. Deprecation warnings can be disabled by setting deprecation_warnings=False in ansible.cfg.
[DEPRECATION WARNING]: Param 'host' is deprecated. See the module docs for more information. This feature will be removed in version 2.9.
Deprecation warnings can be disabled by setting deprecation_warnings=False in ansible.cfg.
We recommend using the new connection types network_cli and netconf (see below), using standard Ansible connection properties, and setting those properties in inventory by group. As you update your playbooks and inventory files, you can easily make the change to become for privilege escalation (on platforms that support it). For more information, see the using become with network modules guide and the platform documentation.
Adding persistent connection types network_cli and netconf
Ansible 2.5 introduces two top-level persistent connection types, network_cli and netconf. With connection: local, each task passed the connection parameters, which had to be stored in your playbooks. With network_cli and netconf the playbook passes the connection parameters once, so you can pass them at the command line if you prefer. We recommend you use network_cli and netconf whenever possible.
Note that eAPI and NX-API still require local connections with provider dictionaries. See the platform documentation for more information. Unless you need a local connection, update your playbooks to use network_cli or netconf and to specify your connection variables with standard Ansible connection variables:
OLD In Ansible 2.4
---
vars:
cli:
host: "{{ inventory_hostname }}"
username: operator
password: secret
transport: cli
tasks:
- nxos_config:
src: config.j2
provider: "{{ cli }}"
username: admin
password: admin
NEW In Ansible 2.5
[nxos:vars]
ansible_connection=network_cli
ansible_network_os=nxos
ansible_user=operator
ansible_password=secret
tasks:
- nxos_config:
src: config.j2
Using a provider dictionary with either network_cli or netconf will result in a warning.