Per-role fact sheets¶
Five fact sheets first produced by systematic source reading of ussuri-eol, then re-studied against 2023.1-eol:
each Ussuri citation was looked up in the 2023.1 sources and each statement re-checked there. Every FACT carries a
path:line reference relative to the kolla-ansible repository root, the kolla image repository (kolla docker/...),
ansible-collection-kolla (ansible-collection-kolla roles/...) or the Ansible source (ansible-core 2.14.18
lib/ansible/...). Anything these sources do not establish is labelled NOT AVAILABLE.
| Sheet | Scope |
|---|---|
| A — infra / load balancing / OVN | loadbalancer (haproxy, keepalived and proxysql; the Ussuri haproxy role, which no longer exists), the 44 per-service loadbalancer.yml includes, loadbalancer-config, haproxy-config, proxysql-config, memcached, chrony (no role in 2023.1), ovn-controller and ovn-db (the Ussuri ovn role), ovs-dpdk, service-cert-copy, certificates.yml |
| B — MariaDB / RabbitMQ / Keystone | Galera bootstrap/recovery/backup and rolling restart; RabbitMQ (incl. outward); Keystone fernet keys and runtime coupling |
| C — Glance / Placement / Cinder / Horizon / Heat | per-role flow, bootstrap on reconfigure, dependencies, restart impact |
| D — Barbican / Octavia / Swift / Manila + compute-side roles | control-plane roles and every role that places containers on compute hosts; templates embedding host lists |
| E — host bootstrap and operational playbooks | bootstrap-servers (in 2023.1 the ansible-collection-kolla roles baremetal, docker, etc_hosts and others: Docker restart conditions, /etc/hosts), prechecks, destroy, prune-images, fact gathering, CLI, tuning docs |
The roles named above as gone are shown by the 2023.1 tree CODE: ansible/roles/ has no haproxy, chrony or ovn
directory, and has loadbalancer, proxysql-config, hacluster, ovn-controller and ovn-db; kolla-host.yml:4-10
applies openstack.kolla.baremetal.
To go from a service or a tag to the right sheet section, use the service index and the tag index.
Note
Key answers
- Eleven corrections to the Ussuri fact sheets, each restated for 2023.1 below (Corrections).
- The one with the widest effect is reversed in 2023.1:
docker stopnow does usedocker_graceful_timeout(60 s), through each container's systemd unit; the Ussuri effective timeout of 10 s no longer applies (F3).
Verification and errata¶
How the sheets were checked:
- Spot checks against the source. The five Ussuri spot checks, re-read in 2023.1 CODE:
- MariaDB restart waves
index % 4: gone in 2023.1. The handlers now only sort hosts into groups (mariadb_bootstrap_restart,mariadb_restart,mariadb_start;roles/mariadb/handlers/main.yml:51-73), and separate plays restart them withserial: "33%"(mariadb.yml:14-40); no% 4remains in the role or the playbook; - glance
log_bin_trust_function_creatorstoggle: still there (roles/glance/tasks/bootstrap_service.yml:2,12,39,49); - Docker restart gate in
bootstrap-servers: moved to ansible-collection-kolla.daemon.jsonand the registry CA notifyRestart docker(ansible-collection-kollaroles/docker/tasks/config.yml:36-47,99-108), which restarts the daemon, or reloads it whendocker_systemd_reloadis set (defaultfalse) (roles/docker/handlers/main.yml:2-15,roles/docker/defaults/main.yml:63); - keepalived priority from inventory index: still there, now in the loadbalancer role
(
roles/loadbalancer/templates/keepalived/keepalived.conf.j2:15); - compute-side memcached lists: still there (
roles/ceilometer/templates/ceilometer.conf.j2:59).
- MariaDB restart waves
- Ansible behaviours that the sheets rely on are cited to the ansible-core 2.14.18 source where it shows them, and are marked NOT AVAILABLE where it does not.
Erratum (sheets A, B and E, and any statement that docker stop honours docker_graceful_timeout: 60).
Ussuri: the effective stop timeout of every kolla_docker stop/restart was 10 s, because the argument spec set
default=10 and a non-None module parameter overrode common_options. 2023.1: the erratum no longer applies
CODE. graceful_timeout and client_timeout are declared without a default (library/kolla_docker.py:303,325), so
the merge takes the common_options values (:351-382): docker_graceful_timeout: 60 and docker_client_timeout:
120 (group_vars/all.yml:105,140,152-153). Containers other than one-shots run under a systemd unit
kolla-<name>-container.service whose ExecStop is docker stop <name> -t <graceful_timeout>
(module_utils/kolla_systemd_worker.py:20-39,63-73); a container without a unit file is stopped by the Docker client
with the same timeout, 10 s only when none is passed (module_utils/kolla_docker_worker.py:436-452).
docker_client_timeout is a working knob too: it is the Docker client timeout
(module_utils/kolla_docker_worker.py:38)
and the time the module waits for a unit to reach its state (module_utils/kolla_systemd_worker.py:59,98,107). A sheet
statement of 60 s therefore matches 2023.1, and one of 10 s describes Ussuri only (F3,
R24).
Erratum (sheet A, "There is no reload path anywhere in scope"). Still holds in 2023.1 CODE. It is true for the
handler path: every handler stops and starts or recreates its container, and kolla_docker has no reload action
(library/kolla_docker.py:263-278). Outside the handlers, Kolla sends one signal: the "Refresh nova scheduler cell
cache" play sends SIGHUP to nova_scheduler on the nova-scheduler hosts. It does so on deploy and reconfigure
runs that select the nova, nova-api or nova-refresh-scheduler-cell-cache tag, or no tag (nova.yml:140-156,
roles/nova/tasks/refresh_scheduler_cell_cache.yml:10).
Erratum (sheet B §3.10, memcached counts). Ussuri: 35 roles had memcache_secret_key in their
templates, 37 had memcache_servers/memcached_servers, and 38 embedded the memcached host list (37 template loops,
plus monasca's monasca_memcached_servers). 2023.1 CODE (grep across roles/*/templates): 29 roles have
memcache_secret_key, 32 have memcache_servers/memcached_servers, and 33 embed the memcached host list, all
by looping over groups['memcached'] in their templates (this includes horizon, whose CACHES.LOCATION uses the loop,
roles/horizon/templates/local_settings.j2:177-182). The monasca role keeps only defaults/ and tasks/ and no
longer builds a memcached list. The list still includes neutron.conf (roles/neutron/templates/neutron.conf.j2:125),
which the neutron agents on compute hosts also read, so a memcached membership change reaches computes.
Erratum (sheet A, "tags: always on the import_role: haproxy-config in every loadbalancer.yml"). In 2023.1
there are 44 loadbalancer.yml files, and each imports loadbalancer-config (not haproxy-config) with
tags: always; loadbalancer-config then imports haproxy-config and proxysql-config under conditions
(roles/loadbalancer-config/tasks/main.yml:2-8). 43 do it directly (e.g. roles/keystone/tasks/loadbalancer.yml:2-7;
ceph-rgw writes role: instead of name:, roles/ceph-rgw/tasks/loadbalancer.yml:2-7).
roles/nova-cell/tasks/loadbalancer.yml reaches it through a chain: import_tasks: proxy_loadbalancer.yml, then a
looped include_tasks: cell_proxy_loadbalancer.yml, then import_role: loadbalancer-config
(roles/nova-cell/tasks/cell_proxy_loadbalancer.yml:3-12), with tags: always at every level. The effect is the same
CODE (Tags §2.3).
Erratum (image-side statements, kolla read later). The kolla image sources were read after the Ussuri sheets were
written. Re-read in kolla 2023.1-eol CODE:
- sheet A: keepalived's
/check_alive.shdoes query the haproxy stats socket (kolladocker/keepalived/check_alive.sh:37). The haproxy image ships a reload helper (kolladocker/haproxy/haproxy/ensure_latest_config.sh:5-12) that no file under kolla-ansible 2023.1ansible/calls; - sheet B: the broker's
.erlang.cookiefile is written only by the RabbitMQ bootstrap container, never on a normal broker start (kolladocker/rabbitmq/rabbitmq/extend_start.sh:7-18; the image moved fromdocker/rabbitmq/); - sheet C: the placement bootstrap container also runs
db online_data_migrations, on every deploy and reconfigure (kolladocker/placement/placement-api/extend_start.sh:5-9). The glance bootstrap addsdb_load_metadefs(kolladocker/glance/glance-api/extend_start.sh:5-9). The heat bootstrap creates the heat domain and user only when the domain is absent (kolladocker/heat/heat-api/extend_start.sh:5-17). Horizon runscollectstaticonly when its settings hash is missing or differs, a dashboard plugin is added or removed, orFORCE_GENERATE=yes(kolladocker/horizon/extend_start.sh:32-46,248-293; Kolla images §5.2); - sheet D: swift config and rings are copied into a container only at its start (kolla
docker/base/start.sh:11,docker/base/set_configs.py:535-551). Onlyswift_rsyncdfixes/srv/nodeownership at start (kolladocker/swift/swift-rsyncd/extend_start.sh:3).
In the same Ussuri pass, sheet E §11 was corrected: like
reconfigure, upgrade overrides a user -e kolla_serial. This still holds in 2023.1 CODE: a user -e is added to
EXTRA_OPTS while the options are parsed (tools/kolla-ansible:360-361), and upgrade and reconfigure then append
-e kolla_serial=${ANSIBLE_SERIAL} after it (:508, :517; command line at :587). Statements that cite a kolla path
(kolla docker/..., relative to the root of the kolla image repository at 2023.1-eol) are CODE. Other image-side
statements in the sheets are NOT AVAILABLE. See Kolla images.
Context the sheets do not repeat. The Ussuri sheets were written before the Ansible source was read for looped
tasks, and this still applies in ansible-core 2.14.18 CODE: when any item changes, Ansible notifies the handlers of
every executed loop item. Every executed item carries the task's notify
(lib/ansible/executor/task_executor.py:804-805),
skipped items return before it is set (:453-455), the task counts as changed when any item changed (:124-128), and
the strategy then notifies the handlers named by each item (lib/ansible/plugins/strategy/__init__.py:614-621).
Wherever a sheet says "restarts container X if its config changed", read it as "restarts all containers of that role
on that host if any of them changed" (see Tags §2.4).
Corrections¶
The Ussuri fact sheets were produced by parallel source-reading passes and spot-checked (MariaDB restart waves, glance
log_bin_trust_function_creators, the Docker restart gate, keepalived priority, compute-side memcached lists). Eleven
corrections applied to them. All are listed with their evidence, and their 2023.1 status, under Verification and
errata, above:
- Fact sheets A, B and E
(and any sentence elsewhere) stating that
docker stopusesdocker_graceful_timeout= 60 s were wrong for Ussuri, where the effective timeout was 10 s. In 2023.1 the 60 s statement is right anddocker_client_timeoutis a working knob (F3). A statement of 10 s, such as the one fact sheet C made for Ussuri, does not hold for 2023.1. - Fact sheet A, "no reload path anywhere": true for handlers, but
nova_schedulerreceives aSIGHUPon everydeploy/reconfigurethat runs thenova,nova-apiornova-refresh-scheduler-cell-cachetag (or no tag) and whose--limitincludes thenova-schedulerhosts (nova.yml:140-156). Still holds in 2023.1. - Fact sheet B §3.10, memcached counts: in 2023.1, 29 roles use
memcache_secret_key, and 33 embed the memcached host list (Ussuri: 35 and 38). - Fact sheet A,
tags: alwaysonimport_role: haproxy-config"in everyloadbalancer.yml": in 2023.1 the import is ofloadbalancer-config; 43 of 44 files do it directly; nova-cell reaches it through a chain of includes. - Fact sheet B, "Graceful stop" row: it stated 60 s, corrected to 10 s for Ussuri; 60 s is the 2023.1 value (F3).
- Fact sheet B, Erlang cookie: the kolla image writes
.erlang.cookieonly in theKOLLA_BOOTSTRAPcontainer, which runs only on a new volume, not at every start (still so in 2023.1,roles/rabbitmq/tasks/bootstrap.yml:10-26). The sheet's former claim of a partition during the rolling recreate rested on that wrong premise; the corrected behaviour is in the second rabbitmq row of Known bugs and limitations. - Fact sheet C, four image-side statements, checked against the kolla
sources and still true in
2023.1-eol(F15, Q14 and Kolla images §4, §5.2):- Placement online data migrations also run on every deploy/reconfigure, inside
bootstrap_placement, not only on upgrade. - The heat domain admin is created only when
heat_user_domainis absent, so a changed password is not re-applied. - Horizon's collectstatic/compress is conditional (settings hash changed or missing, a plugin toggled, or
FORCE_GENERATE=yes), not run on every start. - The glance bootstrap also runs
glance-manage db_load_metadefs.
- Placement online data migrations also run on every deploy/reconfigure, inside
- Fact sheet E, "For every other command,
-e kolla_serial=...works":upgradeoverrides it likereconfiguredoes (tools/kolla-ansible:508,517); useANSIBLE_SERIAL. Still holds in 2023.1.
Related¶
- Technical findings — F3, the stop timeout (no longer ignored in 2023.1), and F15, what the bootstrap containers run per image
- Kolla images — the kolla image sources behind the image-side errata
- Known bugs and limitations — implementation limitations and
bugs found in
2023.1-eol, among them the corrected RabbitMQ behaviour - Tags §2.4 — loop-notify, the context the sheets do not repeat
- What is not known — everything marked NOT AVAILABLE, by topic