| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This brings support for multiple backups pointing to the same borg
server, different repositories, from the same client. We support this by
parsing the SSH authorized key file first and appending allowed
repositories to allow independent definitions of the borg targets.
So that the decryption_keys do not clash we include the repository name
as well as the host in the key. This will lead to new keys being created
for existing hosts in the new format.
Old Format: {{ combine({inventory_hostname: borg_keys.stdout}) }}
New Format: {{ combine({(inventory_hostname ~ '_' ~ borg_repo_name): borg_keys.stdout}) }}
If upgrading from an older version that used just the hostname as the
key, your existing `decryption_keys.yml` can be manually removed once
the new format is also added.
|
| |
|
|
|
|
|
|
| |
- Add test systems for client/server setup
- Add tests for systemd service and timer files
- Add tests for SSH connectivity
- Add tests for security configurations
- Add tests for multi-instance backup scenarios
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The blockinfile marker was only keyed by borg_server_host_url, causing
backup blocks to overwrite each other when running the role multiple
times on the same host with different repositories.
This change includes borg_repo_name in the marker, allowing multiple
backup blocks to coexist in the base script. Now running the role
twice with different borg_repo_name values creates separate blocks
instead of overwriting.
This ensures idempotent behavior and allows the base script to
accumulate all backup targets as documented.
Before: marker for "borg-server" (overwrites on second run)
After: marker for "borg-server/configs" and "borg-server/home-data"
|
| | |
|
| |
|
|
|
| |
If passphrase contains special characters like (!, $, &, ^, *, @), this needs to be escaped.
Co-authored-by: Colin Wilk <colin@wilk.cx>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Systemd gives us the ability to monitor backup job status using existing
monitoring solutions (node exporter) and allows us greater control over
the scheduling of the backup jobs.
This introduces a breaking change that requires users to manually remove
the old repositories from the clients and redeploying them with the
role. You will have to remove the Cron job that was created by the
Ansible script, everything else will be overwritten with a run from the
newer version.
- name: Remove backup cron jobs
ansible.builtin.cron:
name: BORG (Application level backups)
state: absent
become: true
- name: Remove env for backup cron job
ansible.builtin.cron:
name: BORG_PASSPHRASE
env: true
state: absent
become: true
Performing manual migrations on the Borg server is not required.
We now additionally support multiple Borg repositories per client host
using the `borg_backup_argument` variable.
Signed-off-by: Colin Wilk <colin.wilk@tum.de>
|
| |
|
|
|
|
|
| |
Change single logic condition with `and` string into two logic
conditions in a condition list.
Signed-off-by: Colin Wilk <colin.wilk@tum.de>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fix linting errors that came up with the new rules
Main changes:
* Enforce YAML document start headers
* Enforce spacing with commas, colons and hyphens
* Enforce indentation
* Enforce use of quotes only then required
* Enforce use of single quotes over double quotes
* Enforce use of true/false over yes/no and derivatives
* [..] Other minor rules that were already followed
Signed-off-by: Colin Wilk <colin.wilk@tum.de>
|
|
|
Signed-off-by: Colin Wilk <colin.wilk@tum.de>
|
