Critical fix for VPS/server deployments:
- Gateway now uses user-level systemd (~/.config/systemd/user/)
- Agent monitor uses system-level systemd (/etc/systemd/system/)
- Added loginctl enable-linger requirement for persistence
- Set XDG_RUNTIME_DIR and DBUS_SESSION_BUS_ADDRESS env vars
- Updated deploy.sh with proper environment setup
- Enhanced health check to verify linger and runtime dir
- Updated agent-monitor.js with reliable gateway detection
Documentation:
- Added comprehensive systemd troubleshooting guide to MEMORY.md
- Documented user-level vs system-level service architecture
- Created deployment checklist for new servers
- Added best practices for multi-agent deployments
Files changed:
- systemd/openclaw-gateway-user.service (new)
- systemd/openclaw-agent-monitor.service (updated)
- deploy.sh (complete rewrite of service management)
- agent-monitor.js (improved gateway status checks)
- MEMORY.md (added systemd troubleshooting guide)
- Added systemd services for system-level persistence (gateway + monitor)
- Enhanced agent-monitor.js with auto-healing and Telegram notifications
- Created deploy.sh for one-click deployment and management
- Updated CORE_INDEX.md with complete architecture documentation
- Updated MEMORY.md with implementation details and usage guide
- All memory files now tracked in git for version control and rollback
Features implemented:
✓ System-Level: Services auto-start on boot, survive logout/reboot
✓ Auto-Healing: Crash detection, auto-restart with rate limiting
✓ Multi-Layer Memory: Core (CORE_INDEX.md) + Long-term (MEMORY.md) + Daily (memory/)
✓ Git Rollback: ./deploy.sh rollback / rollback-to <commit>
✓ Telegram Notifications: Alerts on stop/error/restart events
Eason (陈医生)