Container Health Watcher

donna-healthwatch is an independent sidecar that watches the Donna container stack and posts to the Discord debug channel when a container goes unhealthy/down or recovers (edge-triggered — one message per transition).

What it watches

All containers named donna-* plus caddy (configurable). Immich and photo-curator are intentionally excluded.

Reciprocal monitoring

The watcher writes a heartbeat file each poll cycle to a shared volume (/mnt/donna/healthwatch/heartbeat). The orchestrator reads it (read-only, no Docker socket) and alerts to the same channel if it goes stale — so the watcher itself is watched. The only uncovered case is both processes down at once, which generally means the host is down.

Configuration (docker/.env)

Var	Default	Where
DISCORD_BOT_TOKEN	—	sidecar (reused from orchestrator)
DISCORD_DEBUG_CHANNEL_ID	—	sidecar
HEALTHWATCH_POLL_SECONDS	30	sidecar
HEALTHWATCH_WATCH_PREFIX	donna-	sidecar
HEALTHWATCH_WATCH_EXTRA	caddy	sidecar
HEALTHWATCH_STALE_SECONDS	90	orchestrator

Operating

• Logs: docker logs donna-healthwatch
• Restart: docker restart donna-healthwatch
• A poll_failed log line means the Docker socket was unreachable that cycle; the watcher skips the cycle (and does not heartbeat) and retries.