Summary#

Build a restart tracking screen in Mission Control so Maxime can see when/why the server restarted (planned rebuild, crash, watchdog trigger, systemd restart).

Existing Infrastructure#

1. scripts/record-startup.js already logs to /tmp/mc-restart-log.jsonl with: ts, pid, ppid, parent process, reason (systemd/watchdog-restart/manual/unknown), prevUptimeSec, bootId
2. start.sh line 19 calls node scripts/record-startup.js at every boot
3. lib/activity-log.ts already defines system.restart action type and has recordActivity() for structured NDJSON logging
4. /api/activity-log POST creates activity events, GET queries them
5. /activity?view=audit already displays all activity events including restarts with 🔄 emoji
6. lib/metrics.ts has time bucketing utilities

What Needs Building#

Phase 1 — Data Pipeline#

• scripts/record-startup.js: after writing to /tmp/mc-restart-log.jsonl, also call recordActivity() (via child_process or direct import) to log a system.restart event to the structured activity log
  • actor_type: "system", actor_id: "mission-control"
  • entity_type: "service", entity_id: "mission-control"
  • metadata: { reason, prevUptimeSec, bootId, build_id, build_id_changed }
• start.sh watchdog trigger_restart(): log a system.restart with reason="watchdog-$REASON" before calling systemd restart

Phase 2 — Visualization#

• New page: app/restarts/page.tsx — Restart History dashboard
  • KPI cards: total restarts (24h/7d), last restart time, crash rate %, current uptime
  • Timeline chart: restart events over time, color-coded by reason (crash=red, systemd=orange, watchdog=yellow, manual=blue, planned=green)
  • Event list table: ts, reason, prevUptimeSec, bootId, with quick links to logs for that time window
• New component: components/RestartsClient.tsx (follow MetricsClient pattern)
• API endpoint: app/api/restarts/route.ts — loads data from activity log (system.restart events)
  • Supports range: 24h, 7d, 30d
  • Returns: events[], summary KPIs

Phase 3 — Navigation#

• Add /restarts to lib/navItems.ts with RotateCcw icon

Phase 4 — Gateway Restart Tracking (🆕 added 2026-05-10)#

The OpenClaw gateway (openclaw-gateway.service) already logs rich restart events to /tmp/openclaw/openclaw-YYYY-MM-DD.log. These are NDJSON lines with a message field containing restart-related events:

• Add lib/gateway-restart-parser.ts — reads gateway log files, extracts restart events by pattern-matching the message field, deduplicates by timestamp
• Gateway restart events get a source: "gateway" tag to distinguish from MC restarts
• The /api/restarts endpoint combines MC activity-log events + gateway log events, sorted by timestamp
• RestartsClient shows source badge (🟢 MC / 🔵 Gateway) for each event
• Timeline chart color-codes by category: gateway-restart=purple, channel-restart=orange, hot-reload=yellow, mc-restart=blue, crash=red, planned=green
• KPI cards show counts separately (X MC restarts, Y gateway restarts, Z channel restarts)
• Gateway log files rotate daily — parser reads last 7 days
• Tests for the parser
• Each gateway restart event links to the relevant line in the gateway log

Files to Touch#

• scripts/record-startup.js (Phase 1)
• start.sh (Phase 1 — watchdog trigger)
• app/restarts/page.tsx (new)
• components/RestartsClient.tsx (new)
• app/api/restarts/route.ts (new)
• lib/navItems.ts (Phase 3)
• lib/gateway-restart-parser.ts (NEW — Phase 4)
• tests/gateway-restart-parser.test.ts (NEW — Phase 4)

Acceptance Criteria#

• Restart events are recorded to activity log (visible in /activity?view=audit)
• /restarts page shows restart history with timeline visualization
• Color coding distinguishes crash vs planned vs watchdog vs manual
• KPIs show restart frequency and current uptime
• Gateway restarts are parsed from gateway logs and shown alongside MC restarts
• Each gateway restart event links to the relevant gateway log line
• Timeline distinguishes gateway, channel, config-reload, and MC restarts
• Build passes, no regressions
• Vera reviews the code after submission