[ideaship]

This removes 19 defaults that nothing reads and fixes one misnamed variable.

The findings come from two passes:

• a manual consumer sweep across the osism org (private repos included), upstream kolla-ansible, ceph-ansible, and the external roles;
• the defaults_orphan drift check:
  • Add a drift check for unread osism/defaults variables release#2811

Each commit covers one file and traces the git history of every key it removes: which reader went away, when, and in which commit.

• kolla (099-kolla.yml):
  • certificates_dir: upstream renamed it to kolla_certificates_dir in Ussuri;
  • keystone_token_provider: upstream hard-codes fernet since Zed.
• infrastructure (099-infrastructure.yml):
  • the configure_* helper section, whose role was split up and archived in 2020;
  • cephclient_cluster_fsid;
  • opentstackclient_groupname, misspelled since it first appeared.
• ceph (099-ceph.yml):
  • bootstrap_dirs_owner and bootstrap_dirs_group, gone from ceph-ansible since 2018 and 2020;
  • ceph_share_directory, never read;
  • osd_scenario, gone from ceph-ansible's roles since stable-4.0. It is only passed to a test playbook by tox-update.ini.
• k3s (099-k3s.yml):
  • cilium_cli. Its gate went with k3s: remove cilium installation osism-kubernetes#241.
• generic (099-generic.yml):
  • docker_configure_storage_driver;
  • grub_hidden_timeout, dead since the debops 2018 grub redesign;
  • fluentd_host, a reverse alias of rsyslog_fluentd_host. The alias itself in the rsyslog role is removed separately:
    • rsyslog: drop reverse fluentd aliases ansible-collection-services#2180
  • sysdig_configure_repository, whose role was removed in 2024.
• interfaces (099-interfaces.yml):
  • bifrost_network_address_family. kolla-ansible never reads it. The copy in 001-bifrost.yml stays, because that layer mirrors upstream.
• ceph images (002-images-ceph.yml):
  • ceph_docker_tag. Nothing ever read it, and its value had a stray }. The cephadm plays read ceph_docker_image_tag.
• The monitoring k3s label (monitoring/000-defaults.yml):
  • osism-kubernetes collects node labels with merge_variables matching ^k3s_add_labels__.+$.
  • The monitoring group's variable was k3s_add_labels_monitoring, with one underscore, so monitoring nodes never got node-role.osism.tech/monitoring-plane=true.
  • It is renamed to k3s_add_labels__monitoring. Nothing selects on that label yet, so no workload moves.

Kept on purpose:

• grafana_server_group_name and iscsi_gw_group_name stay, because ceph-ansible stable-7.0, which quincy pins, still reads them.
• keystone_admin_port and keystone_admin_listen_port are dead upstream, but an environment overlay still templates them. They are handled in a separate change.

Run against this branch, the defaults_orphan check (osism/release#2811) reports no findings. This PR and osism/release#2811 should land together; otherwise the nightly drift job fails on these findings.

🤖 Generated with Claude Code