A simple bugfix on my personal server’s Ansible playbook

To establish context, I have a Terraform script that automatically creates a new Droplet on DigitalOcean, running Ubuntu 22.04. This droplet can then be set up to run nginx and certbot by a simple Ansible playbook.

Also unrelated but I have a small script that reads the output log of Terraform to get the IP address of the droplet:

const fs = require('fs');

fs.readFile('./terraform.tfstate', 'utf8', (err, data) => {
    if (err) {
        console.log(`Error reading file from disk: ${err}`);
    } else {
        const tfstate = JSON.parse(data);
        if (tfstate.resources.length == 0) {
            console.log("no IP yet");
            return;
        }
        console.log(tfstate.resources[0].instances[0].attributes.ipv4_address)
    }
});

Everything being orchestrated by a big Makefile, it just has to read the IP from this script and add it to the Ansible inventory.

The bug

I modified a few nginx configuration files and some files were renamed in the process. As destroying and recreating a droplet costs time, I often run my ansible playbook several time on a single droplet.

Which means that the old config files were still there, and, because I leverage certbot’s ability to automatically modify nginx config files to test domain ownership, I ended up breaking those tests.

To fix this I added a step that removes every old config files.

- name: Remove every old Nginx enabled configs
  ansible.builtin.file:
    state: "{{ item }}"
    path: "/etc/nginx/sites-enabled"
    owner: 0 # root
    group: 0
    mode: '0755' # i checked that those were the default permissions for these folders
  with_items:
    - absent # destroys
    - directory # recreate
- name: Remove every old Nginx available configs
  ansible.builtin.file:
    state: "{{ item }}"
    path: "/etc/nginx/sites-available"
    owner: 0
    group: 0
    mode: '0755'
  with_items:
    - absent
    - directory