[shipmill-romamo]

Plan

• Serves: G-3
• Files: config/schemas/v1/component.schema.json, sdk/src/cloudfall/domain.py, sdk/src/cloudfall/inventory.py, sdk/src/cloudfall/lifecycle.py, engine/ansible/roles/cloudfall_deploy/meta/argument_specs.yml, engine/ansible/roles/cloudfall_deploy/tasks/{main,restart,rollback,health}.yml, sdk/tests/test_command_health_check.py, engine/tests/test_deploy_role.py, engine/tests/test_ansible_inventory.py, changelog.d/161-command-health-check-schema.md
• Order: failing tests first (config validate accepts/refuses the command check, the inventory parses and renders it, deploy/restart/rollback plans name it, the engine refuses it until Gate a deploy on the command health check and receipt it (S-009-2..9, 11, 12) #164); then the schema oneOf branch, HealthCheckType.COMMAND and its value objects, the parser, the preview effects, the engine refusal; then the changelog fragment
• Proof: S-009-1, S-009-13 (test_s009_1_*, test_s009_13_*)

Summary

Adds the config contract of spec 009's command health check; the engine running it is #164.

• config/schemas/v1/component.schema.json: a third healthCheck oneOf branch, type: command with a required command (non-empty list of non-empty strings, NUL refused as CommandArgv refuses it), timeoutSeconds (1 to 300) and attempts (1 to 20), additionalProperties: false. The http and none branches are unchanged
• sdk/src/cloudfall/domain.py: HealthCheckType.COMMAND, and the value objects HealthCheckTimeout (1..300) and HealthCheckAttempts (1..20)
• sdk/src/cloudfall/inventory.py: ComponentCommandHealthCheck (CommandArgv, HealthCheckTimeout, HealthCheckAttempts), an optional ComponentHealthCheck.command required exactly for type: command, its parser and as_dict (so the rendered Ansible inventory carries it as declared)
• sdk/src/cloudfall/lifecycle.py: deploy, restart and rollback without --yes add "; its command health check <command> runs on each server after its restart" to wouldRun. The plan JSON gains no field
• Engine, carry-through only: the cloudfall_deploy argument spec accepts type: command and its command key; main.yml, restart.yml, rollback.yml and health.yml start with a task that fails for a command check ("Component X declares a command health check, which this engine does not run yet (Gate a deploy on the command health check and receipt it (S-009-2..9, 11, 12) #164); nothing was changed"), so until Gate a deploy on the command health check and receipt it (S-009-2..9, 11, 12) #164 lands no deploy, restart or rollback of such a component passes an unrun gate. cloudfall health reports it unhealthy with that message. Gate a deploy on the command health check and receipt it (S-009-2..9, 11, 12) #164 replaces these tasks with the check itself

Test plan

• New sdk/tests/test_command_health_check.py: S-009-1 (accepts, bounds, each malformed case exits CONFIG_INVALID/schema_validation_failed, value objects, Q-3 shapes of http/none unchanged) and S-009-13 (the three plans exit 0, name the check and its command, run no ansible-playbook, and carry the same keys as for http/none)
• engine/tests/test_ansible_inventory.py: a command check renders as declared; engine/tests/test_deploy_role.py: each entry point refuses a command check first and only a command check, and the argument spec accepts all three types
• Ran: uv run ruff check ., uv run mypy, uv run cloudfall config validate --project config/examples, ansible-lint engine/ansible, and pytest on the touched and neighbouring files (test_command_health_check, test_validation, test_inventory, test_lifecycle, test_release_command_plan, test_ansible_inventory, test_deploy_role, test_release_command, test_deploy_templates, test_artifact_build_steps, test_catalog, test_ansible_reader): all green. The full suite is left to CI

Fixes #161
Part of #152

🤖 Generated with Claude Code