Command quick lookup by tag¶
Impact Restarts on drift Writes config Pulls images
One row per command: the hosts it touches with the 2024.1 multinode example inventory, the containers that can
restart, the one-shot containers it runs and what scales its work. Fifteen rows are -t <tag> reconfigure (the
-t chrony row is kept to record that the tag still selects nothing in 2024.1; -t letsencrypt and -t etcd are new
rows); the last four are reconfigure without a tag, deploy-containers, genconfig and pull. Each row is checked
against the 2024.1 code; Tags §3 gives the full detail.
Still holds in 2024.1: no row brings the common containers along except -t common and reconfigure without a tag.
common is a play of its own, tagged common (site.yml:88-103), and no role declares a dependency (there is no
roles/*/meta directory in 2024.1). "haproxy" below is the loadbalancer group (the network hosts in multinode,
inventory/multinode:84-85), reached through the per-service include_role … tasks_from: loadbalancer of the
loadbalancer play, which runs only with enable_haproxy and for deploy, reconfigure, upgrade and config
(site.yml:105-326, condition at :324-326). Every play now carries max_fail_percentage, default 100, which does not
change what a row touches (Write config first, apply in controlled batches
§1.3).
No row exists for the services removed in 2024.1 (freezer, monasca, murano, sahara, senlin, solum, vitrage, kafka,
storm, zookeeper, outward RabbitMQ): their roles are gone from ansible/roles/ and no play selects them, so their tags
run only the always plays (Tags §3) CODE.
The table follows the conventions of the command sections, which also say what the examples assume: "Can restart" is the Restarts convention, containers that can restart, only if drift is detected. "Typical duration driver" names what the amount of work grows with, read from the hosts and tasks a run covers; no run was timed, so durations themselves are NOT AVAILABLE.
| Command | Hosts touched (default inventory) | Can restart | One-shot containers | Typical duration driver |
|---|---|---|---|---|
-t keystone reconfigure |
haproxy, control (keystone group, inventory/; play site.yml:483-495) |
keystone, keystone_ssh, keystone_fernet (roles/), haproxy |
bootstrap_keystone (db sync, kolla docker/), run_once (roles/); bootstrap_keystone_fernet only when no keystone_fernet container is running (:52-75) |
small |
-t glance reconfigure |
haproxy, glance-api (control, inventory/; play site.yml:544-556) |
glance-api (and glance-tls-proxy when enabled, roles/), haproxy |
bootstrap_glance (db sync + metadefs load, kolla docker/), between two kolla_toolbox calls that set and reset the MariaDB global log_bin_trust_function_creators, unless use_preconfigured_databases (roles/) |
small |
-t placement reconfigure |
haproxy, control (inventory/; play site.yml:593-605) |
placement-api, haproxy | db sync + online data migrations in one bootstrap_placement container (roles/; kolla docker/) |
small |
-t nova reconfigure |
haproxy, control, all computes | all nova incl. libvirt, haproxy | 6 runs (those of nova-api and nova-cell below) when cell0 and the cell are unchanged |
number of computes |
-t nova-api reconfigure |
haproxy, control (nova.yml:29-65, :103-122, :152-172) |
nova-api, scheduler (super-conductor only with enable_cells), haproxy |
3: API DB sync (roles/, bootstrap_service.yml:2-21), map_cell0 and a list_cells for cell0 (map_cell0.yml:3-30; roles/); update_cell for cell0 and a second API DB sync only on change (map_cell0.yml:32-67) |
small |
-t nova-cell reconfigure |
haproxy (console-proxy frontends, site.yml:259-265), control (conductor, consoles), all computes (nova.yml:125-147) |
conductor, consoles, compute, libvirt, ssh; haproxy if a console-proxy frontend changed | 3: cell DB sync (roles/), list_cells ×2: cell lookup (create_cells.yml:2-4) and discovery (discover_computes.yml:4-5, once per batch that holds computes, wait_discover_computes.yml:87-90); create_cell/update_cell only when the cell is missing or its URLs changed (create_cells.yml:6-55); discover_hosts itself is an exec into nova_conductor (discover_computes.yml:14-18) |
number of computes |
-t neutron reconfigure |
haproxy, control, network, all computes (play hosts include compute and manila-share, site.yml:671-693) |
server, all agents, haproxy | bootstrap_neutron (neutron-db-manage upgrade head for neutron and each enabled subproject). Still holds in 2024.1: it runs on every deploy and reconfigure, whatever neutron_enable_rolling_upgrade (roles/, deploy.yml:20, bootstrap.yml:38, bootstrap_service.yml:2-21, no when) |
number of computes |
-t openvswitch reconfigure |
network, all computes (openvswitch group, inventory/; play site.yml:611-624, only while enable_ovs_dpdk is off) |
openvswitch_db/vswitchd (roles/); a changed system-id, hostname or hw-offload restarts vswitchd (post-config.yml:22-23) |
– | number of computes |
-t haproxy reconfigure |
haproxy (site.yml:105-120) |
haproxy, keepalived, proxysql when enabled (backup hosts before the master, roles/); new: haproxy_ssh with enable_letsencrypt (roles/, handlers/) |
– | small |
-t letsencrypt reconfigure (new) |
only with enable_letsencrypt (default "no", group_vars/): haproxy (letsencrypt include, site.yml:208-212) and the letsencrypt group, a child of loadbalancer (inventory/; play site.yml:328-340) |
letsencrypt_webserver, letsencrypt_lego (roles/), haproxy if its frontend changed; not haproxy_ssh, which belongs to the loadbalancer play |
– (roles/) |
small |
-t chrony reconfigure |
none: still no chrony role or play in 2024.1 (only the always-tagged plays run) |
– | – | – |
-t common reconfigure |
all hosts of the common group (site.yml:88-103; inventory/) |
fluentd, cron, kolla_toolbox (all three on drift, loop-notify, roles/) |
–; new in 2024.1: ensure_image pulls the fluentd image when it is absent (roles/) |
number of hosts |
-t mariadb reconfigure |
haproxy, mariadb | mariadb (restart play with serial: "33%": one host per batch with 3 nodes, mariadb.yml:18-34; ansible-core 2.16.20 lib/), clustercheck (enabled with enable_haproxy, roles/), haproxy |
– | Galera sync |
-t rabbitmq reconfigure |
haproxy, rabbitmq (no outward RabbitMQ any more: rabbitmq.yml:3-70 has three plays) |
rabbitmq one by one (serial: "33%", each node drained first: rabbitmq.yml:30-49, roles/), haproxy |
rabbitmq_version_check on the first RabbitMQ host when the container exists (roles/, version-check.yml:18-36, :75-78); rabbitmq_bootstrap only for a new volume (bootstrap.yml:10-26) |
broker start |
-t etcd reconfigure (new) |
only with enable_etcd (default "no", group_vars/): haproxy (etcd include, site.yml:163-167), etcd hosts (control, inventory/; play site.yml:470-481, no serial) |
etcd: non-leaders in % 4 waves, then the leader (roles/); not batched by ANSIBLE_SERIAL; haproxy if its frontend changed |
–; a new cluster or a member without a volume is bootstrapped by starting the etcd container itself (roles/, bootstrap_cluster.yml:2-14) |
cluster size |
reconfigure (no tag) |
all | everything that drifted | db sync of every enabled service, neutron included (see the neutron row), the 6 nova runs, rabbitmq_version_check |
number of hosts × tasks |
deploy-containers -t X |
X's hosts, no haproxy include (site.yml:324-326); no kolla_serial is passed (tools/) |
X (drifted only): the role runs only its container check (e.g. roles/) |
none | small per host |
genconfig -t X |
X's hosts (+haproxy) (tools/, kolla_action=config) |
none: restart handlers skip kolla_action == "config" (e.g. roles/, roles/) |
none; new: -t common can pull the fluentd image (roles/) |
templating |
pull -t X |
X's hosts (tools/) |
none | none | image size × hosts |
Related¶
- Tags §3 — the master table: what each tag executes and on which hosts
- Tags §3.1 — per-tag answers: files, VMs, APIs, networking and storage, safety, prerequisites
- Tag index — each tag with its component page
- Operations catalogue — the operation pages, with their full command lines
- Predict and inspect a run — how to see what will restart before running one of these commands
- Kolla images §4 — what each one-shot container runs
- Compute scaling §2 — the work per compute host, read from the task files
Lumo: ready for the next step? Take your time: you can come back to this page whenever you need it.
Lumo: nicely done! One more page on your journey.