Chapter 10: Automating Linux Administration Tasks

Archiving and restoring files

Backups, releases and log collections all end up as archives. Ansible has one module to pack files on a managed host and another to unpack them, and with a few options both behave predictably on every run. A backup is only useful if you can restore it, so you will also check that a restore gives back the same files.

Create compressed archives on managed hosts with community.general.archive.
Extract archives with ansible.builtin.unarchive, from the control node or from the managed host.
Control what goes into an archive and where its files land.
Prove that a restore gives back the right content, owner and mode.

Two modules, two directions

	`community.general.archive`	`ansible.builtin.unarchive`
Does	Packs files into an archive	Unpacks an archive into a directory
Source	Files on the managed host	An archive on the control node, or on the managed host with `remote_src: true`
Result	An archive file on the managed host	Files in an existing directory on the managed host
Formats	`gz` (default), `bz2`, `xz`, `zip`, `tar`	tar (any of those compressions) and zip

archive is in the community.general collection, which is installed on workstation. unarchive is part of ansible-core. Both need the archiving tools on the managed host, such as tar and gzip, which every RHEL system has.

Create an archive

- name: The release is packaged
  community.general.archive:
    path: /srv/release/
    dest: /tmp/release.tar.gz
    format: gz
    mode: "0644"

path is what to pack: a file, a directory, a list, or a wildcard such as /var/log/*.log.
dest is the archive to create. format chooses the compression; gz gives a .tar.gz file.
mode, owner and group apply to the archive file itself.

The trailing slash decides the layout

`path`	The archive contains	Extracted into `/srv/restored`
`/srv/release`	the directory `release/` and its files	`/srv/restored/release/version.txt`
`/srv/release/`	only the directory's contents	`/srv/restored/version.txt`

Decide where the files must land after a restore, then choose the slash.

Extract an archive

- name: The restore directory exists
  ansible.builtin.file:
    path: /srv/restored
    state: directory
    mode: "0755"

- name: The release is restored
  ansible.builtin.unarchive:
    src: /tmp/release.tar.gz
    dest: /srv/restored
    remote_src: true
    owner: root
    group: root

dest must already exist: unarchive does not create it.
remote_src: true says the archive is already on the managed host. Without it, unarchive looks for src on the control node and copies it across first, which is how you ship a release that you built on the control node.
owner, group and mode apply to the extracted files.

Check the restore, not just the archive

An archive that exists is not proof of anything. Check that the restored files match the originals in content and in access:

[student@workstation ~]$ ssh devops@servera.lab.example.com \
  'cmp /srv/release/version.txt /srv/restored/version.txt && stat -c "%a %U:%G" /srv/restored/version.txt'
644 root:root

cmp prints nothing and succeeds when two files are identical. If you put checks like these in a playbook, add changed_when: false: they only read, so they must not inflate the change count.

The archive `/tmp/site.tar.gz` is already on servera. What must the `unarchive` task include so that Ansible does not look for it on the control node?
You archive `path: /var/www/html/` and extract into `/srv/site`. Where does `index.html` end up?
An `unarchive` task fails because `/srv/site` does not exist. What is the fix?