plugins.filter.swarm package¶

Submodules¶

plugins.filter.swarm.deployed_services module¶

Name the services a docker stack deploy run reported acting on.

class plugins.filter.swarm.deployed_services.FilterModule¶

Bases: object

filters()¶

plugins.filter.swarm.deployed_services.swarm_deployed_services(deploy_result: Any) → list[str]¶

Return the service names from a registered stack-deploy result.

Parameters:: deploy_result – the registered shell result of the deploy.
Returns:: Service names in the order the deploy reported them.

plugins.filter.swarm.restart_condition module¶

Filter swarm_restart_condition: map a compose-style docker_restart_policy to swarm’s deploy.restart_policy.condition.

docker stack deploy does NOT understand compose’s restart: keys (no, always, unless-stopped, on-failure[:N]). It uses a disjoint set of condition: values: none, on-failure, any. Translating between the two used to be inline Jinja in roles/sys-svc-container/templates/deploy.yml.j2 – this filter pulls it out so the call site reads as a single expression and is unit-testable.

MAPPING¶

'no' -> 'none'
One-shot containers (matomo bootstrap, erpnext configurator, shopware init). They exit 0 and must never be respawned.

'on-failure' or 'on-failure:N' -> 'on-failure'
Respawn only on non-zero exit. Swarm has no separate max-restarts knob on the condition itself; the :N suffix is dropped because swarm carries the same idea via restart_policy.max_attempts.

'always' / 'unless-stopped' -> 'any'
Long-running services – the historical default of this deploy template.

anything else (None, empty, typo) -> 'any'
Safe default: long-running services are the most common case and a typo in a role’s docker_restart_policy would otherwise silently produce none, which would skip respawning real services and look like a deploy success.

class plugins.filter.swarm.restart_condition.FilterModule¶