M26-2 Reconcile + watchdog: lazy at boot, no flap event

No labels

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

charles/claude-hooks#589

claude-desktop commented

2026-04-30 19:31:16 +00:00

Collaborator

User story

As a service operator, I want lazy containers to be created at boot but stay stopped, and I want the watchdog to recognise that stopped-lazy is normal, so that the fleet's idle state doesn't flood the event log with false container_stopped flap reports.

Acceptance criteria

Reconcile

container-reconcile.ts reads lifecycle: "hot" | "lazy" from the SQLite agents row.
At reconcileAll() boot pass: hot rows behave as today (start if stopped). Lazy rows ensure container exists with the right config but do not start it. Initial state seeds as Stopped.
reconcileOne() on CRUD applies the same split for create + update.
lifecycle change counts as drift → recreate (so --restart flag is updated).

Restart policy

Hot containers continue using --restart unless-stopped.
Lazy containers use --restart no so the daemon does not undo intentional stops.
container.ts docker run arg builder branches on lifecycle.

Watchdog

container-watchdog.ts reads lifecycle. For lazy + currently-stopped (consult lifecycle module from M26-1), container_stopped event is suppressed.
container_missing (#132 failure mode — not in docker ps -a) still fires for both hot and lazy. Recovery path unchanged.

Tests

Reconcile unit cover: hot row → start, lazy row → exists+stopped, lifecycle change → recreate.
Watchdog unit cover: lazy + Stopped → no event, lazy + Missing → event + recreate, hot + Stopped → event as today.

Out of scope

Lifecycle module itself (M26-1).
Pool selection (M26-3).
Config schema / CRUD surface (M26-4).
SSE / metrics events (M26-5).

References

specs/container-lazy-lifecycle.md §Reconcile changes / §Watchdog changes / §Restart policy.
apps/server/src/infrastructure/container/container-reconcile.ts.
apps/server/src/infrastructure/container/container-watchdog.ts.
Issue #188 — restart-drops-half-the-fleet (existing started action variant).
Issue #132 — watchdog container_missing rationale.

## User story As a service operator, I want lazy containers to be created at boot but stay stopped, and I want the watchdog to recognise that stopped-lazy is normal, so that the fleet's idle state doesn't flood the event log with false `container_stopped` flap reports. ## Acceptance criteria ### Reconcile - [ ] `container-reconcile.ts` reads `lifecycle: "hot" | "lazy"` from the SQLite agents row. - [ ] At `reconcileAll()` boot pass: hot rows behave as today (start if stopped). Lazy rows ensure container *exists* with the right config but do **not** start it. Initial state seeds as `Stopped`. - [ ] `reconcileOne()` on CRUD applies the same split for create + update. - [ ] `lifecycle` change counts as drift → recreate (so `--restart` flag is updated). ### Restart policy - [ ] Hot containers continue using `--restart unless-stopped`. - [ ] Lazy containers use `--restart no` so the daemon does not undo intentional stops. - [ ] `container.ts` `docker run` arg builder branches on `lifecycle`. ### Watchdog - [ ] `container-watchdog.ts` reads `lifecycle`. For lazy + currently-stopped (consult lifecycle module from M26-1), `container_stopped` event is suppressed. - [ ] `container_missing` (#132 failure mode — not in `docker ps -a`) still fires for both hot and lazy. Recovery path unchanged. ### Tests - [ ] Reconcile unit cover: hot row → start, lazy row → exists+stopped, `lifecycle` change → recreate. - [ ] Watchdog unit cover: lazy + Stopped → no event, lazy + Missing → event + recreate, hot + Stopped → event as today. ## Out of scope - Lifecycle module itself (M26-1). - Pool selection (M26-3). - Config schema / CRUD surface (M26-4). - SSE / metrics events (M26-5). ## References - `specs/container-lazy-lifecycle.md` §Reconcile changes / §Watchdog changes / §Restart policy. - `apps/server/src/infrastructure/container/container-reconcile.ts`. - `apps/server/src/infrastructure/container/container-watchdog.ts`. - Issue #188 — restart-drops-half-the-fleet (existing `started` action variant). - Issue #132 — watchdog `container_missing` rationale.

claude-desktop added the

labels

2026-04-30 19:32:05 +00:00

claude-desktop added this to the M26 — Container lazy lifecycle milestone

2026-04-30 19:32:15 +00:00

dev was assigned by code-lead

2026-04-30 21:57:47 +00:00

dev referenced this issue from a commit

2026-04-30 22:25:55 +00:00

feat(agents): M26-2 reconcile + watchdog lazy lifecycle

dev referenced this issue from a pull request that will close it,

2026-04-30 22:26:12 +00:00