3

Ansible lists a nice set of best practices in its documentation. One of the items is:

Always mention the state

For many modules, the state parameter is optional. Different modules have different default settings for state, and some modules support several state settings. Explicitly setting state=present or state=absent makes playbooks and roles clearer.

In my case the main point isn't so much about making playbooks or roles clearer, instead my goal is to manage the presence/absence of items throughout the lifetime of my servers.

Probably many of you have had the issue of installing package xyz as a "bread'n'butter" package on your machines, but a few months later it turns out your workflow has changed and you don't really need xyz anymore. Or it was replaced by xyz-ng ... or what have you ...

So now it's time to express somehow to remove that xyz package consistently across your infrastructure. In my case with Debian/Ubuntu I'd use the apt module with state=absent for the respective packages. No surprise.

However, I'd like to factor this into my roles from the outset. As an example it would be nice to have a dictionary variable (as an example):

packages:
  present:
  - elinks
  absent:
  - lynx

This way the presence and absence would dynamically be managed by variables and I can use when: to cover for an empty list and so on; potentially even by including a task dynamically and passing along a state variable that coincides with the key in the dictionary.

Another case where this must have come up for someone before is the creation of local users, or has everyone moved to directory services? When someone leaves the company I'd like to be able to put that someone on a list populated with usernames to be put into a locked state.

However, since there are established best practices for Ansible in general, I was wondering:

Q: are there best practices about managing the state of "entities" on the configured infrastructure?

2

This can be done with two tasks with loops:

# tasks/packages.yml
- name: pkgs | Ensure Good 
  package:
    name: "{{ item }}"
    state: present
  loop: "{{ packages['present'] }}"

- name: pkgs | Ensure Bad
  package:
    name: "{{ item }}"
    state: absent
  loop: "{{ packages['absent'] }}"

This will loop over all of the items in the list that you define and ensure that they are either present or absent.

You could also use a shorthand:

# tasks/packages.yml
- name: pkgs | Ensure Good 
  package:
    name: "{{ packages['present'] }}"
    state: present

- name: pkgs | Ensure Bad
  package:
    name: "{{ packages['absent'] }}"
    state: absent

This would run faster, since you don't have a list of tasks, but a task with a list, but you would lose some visibility into which packages you're actually ensuring the state of.

Caveats:

Ensure that there is no overlap between the two lists. One way would be to make an assertion on the lists before you run the task:

- name: ensure no overlap between good and bad
  assert:
    that: "{{ packages['present'] | intersect(packages['absent'] }} == [] }}"

Also ensure that the lists are not empty.

To answer your final question:

Q: are there best practices about managing the state of "entities" on the configured infrastructure?

It is usually my practice (I wouldn't call it "best", but take it for what it's worth) to declare the explicit state of a resource. It's quite difficult to do this exhaustively for packages that should be absent - this is something akin to maintaining a "blacklist". One approach might to be start from a policy and implement specific controls in it. IE for ssh config ensure that protocol v1 is absent, etc.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.