Linux – How to merge variables of type hash across different variable files in ansible

ansibleautomationlinuxyaml

I have two YAML variable files which are reused for many playbooks:

# playbook/group_vars/all

settings:
  a: 1
  b: 2

# inventory/group_vars/main.yml

settings:
  c: 3

I want settings to be equal { a: 1, b: 2, c: 3 } in my playbooks.

Instead it equals { c: 3 } because inventory/group_vars/* has
higher priority than playbook/group_vars/all and because hashes are overridden, not merged.

I see the following options:

Define variables separately, using naming conventions and prefixes, instead of using hashes, like so:

# playbook/group_vars/all

settings_a: 1
settings_b: 2

# inventory/group_vars/main.yml

settings_c: 3

Our team does not consider this solution elegant enough.

Use set_fact and/or Jinja's "+" operator

{{ list_common + list_specific }}

This requires changes in many playbooks (we do have many of them) and also looks bad.

—

Are there other options to merge hashes from different files instead of redefining (overriding) them?

What we are trying to achieve is to stop duplicating lists across variable files and define only those parameters which are specific to the group/host.

Best Answer

First of all, list is [1, 2, 3] and you talk about dictionary {a:1, b:2, c:3} or in other words hash.

There is configurable parameter: hash_behaviour.

By default it is replace, but you can set it to merge to achieve desired behaviour.

Related Solutions

Ansible Host Variables – How to Access Host Variable of a Different Host with Ansible

You can access pretty much any inventory facts/variables by doing something like this:

{{ hostvars['foo.example.com']['ansible_eth0']['ipv4']['address'] }}

or, if you want to do it via an index into a group:

{{ hostvars[groups['collectors'][0]]['ansible_eth0']['ipv4']['address'] }}

The big trick is that you need to collect the facts for all the hosts/groups you're interested in. So you would want to modify your playbook that runs against the reporters group to include a no-op (dummy) task that is applied to the collectors group. That will cause Ansible to collect facts about the collectors hosts so that they can be accessed from the reporters group. So you might want to add something like this to the top of your reporters playbook:

- hosts: collectors
  name: Gather facts from collectors
  tasks: [ ]

The empty brackets basically mean that no tasks will be executed, but this will still force Ansible to gather facts about the collectors so that you can then reference them in the tasks that you run against your reporters.

Edit #1

It occurred to me that I should also mention that as of version 1.8 of Ansible, there is a fact-caching feature that is now available. Fact caching relies on a redis server to store facts between playbook runs. With it enabled, one playbook can reference facts that were obtained by another playbook that was run previously. The example the Ansible documentation gives:

Imagine, for instance, a very large infrastructure with thousands of hosts. Fact caching could be configured to run nightly, but configuration of a small set of servers could run ad-hoc or periodically throughout the day. With fact-caching enabled, it would not be necessary to “hit” all servers to reference variables and information about them.

How to do an array or hash merge in Ansible

First of all I had to define two different variables rather than using the same one, e.g. sensu_plugin_install. After giving the second dict a different name like sensu_plugin_install_other, the two arrays were able to merge:

 with_items: "{{ sensu_plugin_install }} + {{ sensu_plugin_install_other }}"

Related Topic