Barbican¶
This page is the starting point for Barbican, the barbican role: the containers barbican_api,
barbican_keystone_listener and barbican_worker (roles/barbican/defaults/main.yml:4,
roles/barbican/defaults/main.yml:28, roles/barbican/defaults/main.yml:36), on the control nodes of the example
inventory, selected by the tag barbican (site.yml:853-867; group barbican is a child of control,
inventory/multinode:125-126). Its DB sync (bootstrap_barbican, run_once,
roles/barbican/tasks/bootstrap_service.yml:2-20) runs on every deploy and reconfigure, and barbican_crypto_key
comes from passwords.yml (etc/kolla/passwords.yml:56, rendered as kek in
roles/barbican/templates/barbican.conf.j2:50 with the default simple_crypto plugin, group_vars/all.yml:1022)
CODE (Component dependencies §3). The lists below say where the study
covers how to operate it, its impact, its dependencies and the evidence, what to check around a run and what stays
open; the service card follows them.
What changed in 2024.1¶
- Database options moved to
[database]CODE:barbican.confno longer setssql_connectionin[DEFAULT]; it renders[database] connection, withconnection_recycle_time,max_pool_sizeandmax_retries = -1(roles/barbican/templates/barbican.conf.j2:18-22). A cloud coming from 2023.1 therefore gets a changedbarbican.confon each barbican host, and the config task notifies the restart of each barbican container (roles/barbican/tasks/config.yml:111-129). What Barbican does with the new options at runtime is NOT AVAILABLE. - New messaging options CODE: with the default quorum queues (
group_vars/all.yml:268),rabbit_quorum_queue = trueis rendered (roles/barbican/templates/barbican.conf.j2:98-107; see F17). Withom_enable_queue_manager(defaultfalse,group_vars/all.yml:269) the three containers mount/dev/shm(roles/barbican/defaults/main.yml:138,roles/barbican/defaults/main.yml:145,roles/barbican/defaults/main.yml:152), andbarbican.confgetsuse_queue_manager(roles/barbican/templates/barbican.conf.j2:87-89) and an[oslo_concurrency]lock path (roles/barbican/templates/barbican.conf.j2:117-120). - Public URL and external frontend CODE: the endpoints are built with the
kolla_urlfilter, the public one onbarbican_api_public_port(group_vars/all.yml:1025-1026), which becomes 443 or 80 withhaproxy_single_external_frontend(defaultfalse) and otherwise stays 9311 (group_vars/all.yml:340-341,group_vars/all.yml:931-932). The external HAProxy entry now carriesexternal_fqdnand the public port (roles/barbican/defaults/main.yml:23-24).host_hrefinbarbican.confis the public endpoint (roles/barbican/templates/barbican.conf.j2:10), so turning the single external frontend on changesbarbican.conftoo. - Container module renamed CODE: the handlers, the check of the containers and the bootstrap container use
kolla_containerinstead ofkolla_docker, with the same actions (roles/barbican/handlers/main.yml:2-48,roles/barbican/tasks/check-containers.yml:2-5,roles/barbican/tasks/bootstrap_service.yml:6-7). - The play CODE: it still takes
serialfromkolla_serial, default0(site.yml:860), and now also setsmax_fail_percentage, default 100 (site.yml:861-864), which never stops the play early (Controlled batches §1.3). - The kolla barbican images are unchanged between the two tags (no difference under kolla
docker/barbican/).
Where to look¶
Operate¶
No operation page gives a command line for the barbican tag, and the tag has no row of its own in Tags
§3 or in the Tag index. What the
study says about running it is in fact sheet D: the plays the tag selects
(§1.1) and the task flow
per action
(§1.3). In
2024.1 the deploy flow is registration, config, check-containers, clone (dev mode only), bootstrap, then the handlers
(roles/barbican/tasks/deploy.yml:2-14); upgrade is config, check-containers, bootstrap_service.yml, then the
handlers (roles/barbican/tasks/upgrade.yml:2-9) CODE.
Dependencies¶
- Dependencies: Component dependencies §3
Evidence¶
- Fact sheet: fact sheet D §1
- Image side: Kolla images
§4 (row
barbican-api), Kolla images §5
Before and after a run¶
- Pre-execution checklist, step 8 (rollback material): previous
/etc/kolla/config,globals.yml,passwords.ymland inventory under git. - fact sheet D
§1.6: the crypto
KEK lives in
passwords.yml(barbican_crypto_key,etc/kolla/passwords.yml:56); this key must be in thepasswords.ymlbackup. - fact sheet D
§1.3:
reconfigure=deploy(roles/barbican/tasks/reconfigure.yml:2), so keystone registration and the DB bootstrap container re-run on every reconfigure CODE. - R2, precautions for enabling or disabling a service that has a logrotate or fluentd
input entry (
enable_Xfor most OpenStack services, e.g. barbican:roles/common/defaults/main.yml:157-158for the fluentd input,roles/common/tasks/config.yml:215for logrotate): treatenable_*changes as fleet-wide changes; roll out with--limitbatches; apply thecommonside deliberately with-t common. In 2024.1, as in 2023.1,td-agent.confandlogrotate.confchange only when thecommonplay runs (an untagged run or-t common,site.yml:88-103), not under-t barbicanCODE. - Reconfigure Nova §3, row
enable_barbican: yes(Barbican not even deployed):nova.confis rewritten everywhere (adds[barbican],roles/nova/templates/nova.conf.j2:217-222,roles/nova-cell/templates/nova.conf.j2:259-264); all nova containers restart, libvirt included on every compute;td-agent.confandlogrotate.confchange too, but fluentd, cron and kolla_toolbox restart only at the next run of thecommonplay, not under-t novaCODE. Cinder renders the same section (roles/cinder/templates/cinder.conf.j2:264-269). - R24, container stop timeout: in 2024.1 the stop goes through the container's systemd
unit, whose
ExecStopis/usr/bin/<engine> stop <name> -t <graceful_timeout>(module_utils/kolla_systemd_worker.py:32), withdocker_graceful_timeout60 s (group_vars/all.yml:139). The first 2024.1 check of a container whose unit 2023.1 wrote reports drift (F16).
Known limits and open points¶
- What is not known: the sources do not establish whether
a
[barbican]section rendered before Barbican exists has any effect before encrypted volumes are used. - Execution flow §2.2: Barbican
comes after Nova, Cinder and Glance in the deployment order; with
enable_barbican, Nova and Cinder render a[barbican]section, and on a first deploy they start configured with it before Barbican exists CODE. - fact sheet D
§1.6,
"Operationally notable", four points: the crypto KEK lives in passwords.yml (NOT AVAILABLE: whether losing or changing
barbican_crypto_keymakes stored secrets undecryptable); p11_crypto: the library path is not mounted into the containers and no task generates the MKEK or HMAC keys (NOT AVAILABLE: what else an HSM setup needs); runningkolla-ansible deploy -t barbicanalone does not run the keystone play, so thebarbican_notificationstopic is not enabled in keystone.conf (NOT AVAILABLE: whether barbican-keystone-listener then receives project-deletion events); copying the policy file usesbecome: true. All four still hold in 2024.1 CODE:etc/kolla/passwords.yml:56;barbican.conf.j2:34-46setslibrary_path,mkek_labelandhmac_label, whilebarbican_library_path(group_vars/all.yml:1023) appears nowhere else underansible/(no task, no volume); the topic is enabled from the keystone role (roles/keystone/defaults/main.yml:191-192);roles/barbican/tasks/config.yml:131-136copies the policy file withbecome: true. - fact sheet D
§7, item
15: Barbican and other bootstraps (octavia, manila, ceilometer, masakari, zun, cyborg) re-run their DB sync container
on every deploy/reconfigure. For Barbican, still holds in 2024.1:
bootstrap_barbicanhas nokolla_actionguard (roles/barbican/tasks/bootstrap_service.yml:2-20) CODE. - F5, "
reconfigure=deploy": a reconfigure re-runs Keystone registration, DB creation/grants and the DB-sync bootstrap containers of most services CODE. - What is not known: fact sheet D has its own NOT AVAILABLE list, and its §1.6 is among the sections that contain NOT AVAILABLE marks.
- Known bugs and limitations, Open questions and Findings have no entry specific to Barbican.
Service card¶
The rows of the card are explained on the Components page.
| Item | Value |
|---|---|
| Play | site.yml:853 |
| Tags | barbican |
| Play host patterns | barbican-api, barbican-keystone-listener, barbican-worker |
| Node types in play (example inventory) | control |
| Node types running its containers | control |
| Containers | barbican_api, barbican_keystone_listener, barbican_worker |
deploy/reconfigure flow (tasks/) |
register → config → check-containers → clone (if) → bootstrap → flush handlers |
| Task files | bootstrap.yml, bootstrap_service.yml, check-containers.yml, clone.yml, config.yml, config_validate.yml, copy-certs.yml, deploy-containers.yml, deploy.yml, loadbalancer.yml, main.yml, precheck.yml, pull.yml, reconfigure.yml, register.yml, stop.yml, upgrade.yml |
| Templates | barbican-api.ini.j2, barbican-api.json.j2, barbican-keystone-listener.json.j2, barbican-worker.json.j2, barbican.conf.j2 |
| Handlers (definition order) | Restart barbican-api container; Restart barbican-keystone-listener container; Restart barbican-worker container |
| Key variables | enable_barbican, barbican_crypto_key (passwords.yml) |
| Role dependencies | none |
Related¶
- fact sheet D §1 — the per-role evidence: plays, services, task flow, config files and dependencies
- Keystone —
-t barbicanalone does not run the keystone play, so thebarbican_notificationstopic is not enabled in keystone.conf - Nova — the compute
nova.confchanges when barbican is enabled - Common: fluentd, cron, kolla_toolbox — enabling a service such as barbican changes
td-agent.confand/orlogrotate.confon every host of thecommonplay, when that play runs - RabbitMQ — quorum queues on by default in 2024.1, rendered into
barbican.conf - Why a small change creates a large workload —
enable_<new service>: yeschanges fluentd/logrotate on every host andnova.confsections on every compute - Components — the other component pages