Add osism get mariadb-backup-host

[ideaship]

osism apply mariadb-backup writes its archives to one MariaDB host, and nothing in osism can say which one before a backup is taken. This adds osism get mariadb-backup-host.

Depends on:

• mariadb: add a playbook reporting the backup host container-image-kolla-ansible#965

That PR adds the playbook this command runs, so it has to be merged first.

Documentation:

• Document osism get mariadb-backup-host osism.github.io#1092

Why the host cannot be looked up today

The mariadb role resolves it from a role default over a shard group built during the play, and the usual override is an extra var from environments/kolla/configuration.yml. None of that is in the inventory:

• osism get hostvars <host> mariadb_backup_host reports the variable as not found;
• osism get hosts -l mariadb sorts alphabetically, so it cannot show the inventory order the default depends on.

Operators are left to probe each MariaDB host for /etc/kolla/mariabackup.

What the command does

It runs the read-only kolla-mariadb-backup-host.yml play on the kolla-ansible worker. The play asks the mariadb role itself: it evaluates the backup task's own condition on every MariaDB host, across all shards, and reports once:

mariadb_backup_host=<host> resolved=<host per shard>

The first field is the host the backup runs on, empty if there is none; the second lists what each shard resolved.

The command dispatches it with kolla.run and publish=False, takes the output from the task result instead of streaming it, and prints only the answer:

• a table by default;
• with --format script, the bare host name for use in shell commands.

It exits non-zero:

• if the play reports no backup host. The error names the resolved candidates: kolla backs up only the default shard and would skip the backup without a word, for example when mariadb_backup_host is overridden to a host outside it;
• if the play reports nothing;
• if the play fails. The play output is not streamed, so the error points at osism apply -e kolla mariadb-backup-host for the full output;
• on timeout. The pending result is then forgotten explicitly; left to interpreter shutdown, Celery prints a traceback.

Why a play and not a local lookup

resolve_in_host_context evaluates expressions in a host's inventory context, but sees neither the role default nor the extra vars. Using it would mean copying kolla's rule into python-osism. The play leaves the resolution to the role, as osism validate already does for its kolla-ansible validators.

Testing

• 7 new unit tests in tests/unit/commands/test_get.py.
• On a 2026.1 cluster, with the playbook placed in the kolla-ansible container and this get.py loaded in osismclient:
  • both output formats printed the host holding /etc/kolla/mariabackup;
  • with a failing play (the playbook removed), it exited 1 with that pointer;
  • with --timeout 1, it exited 1 with the timeout error and no traceback.
• The sharded and no-backup-host reports were verified at the play level, with a second shard simulated through an extra inventory source.

🤖 Generated with Claude Code

[sourcery-ai]

Hey - I've reviewed your changes and they look great!

Sourcery assessment

Approved.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook