UPS-Backed Shutdown Orchestration

Goal

Survive a power loss without corrupting disks or VMs. The UPS signals a controller, which drains guests, parks storage, and brings things back up cleanly.

Topology

• UPS USB/serial into a small controller (Raspberry Pi/thin client/VM).
• NUT server or apcupsd running on that controller.
• NUT clients on: Proxmox hypervisor(s), NAS, key Docker host.

Install & Configure

1. Pick NUT (multi-vendor) or apcupsd (APC-focused); install on controller.
2. Confirm driver works: upsc ups@localhost or apcaccess status.
3. Enable TCP listener; bind to management VLAN; add strong NUT passwords.
4. On clients, install NUT/apcupsd in net-client mode; point to controller IP.

Shutdown Ordering

• Set low battery threshold and kill delay on controller.
• On Proxmox: create hook script to gracefully stop VMs/CTs by priority.
• On NAS: trigger dataset sync and stop SMB/NFS before shutdown.
• On Docker host: stop containers; flush logs if you centralize to Loki.

Restore Flow

• BIOS: power-on after AC loss.
• Bring up NAS first, then hypervisor, then app hosts; delay stacks until storage is ready.
• Check time sync (NTP), then start proxies and tunnels last.

Testing

• Simulate UPS on battery for 2-3 minutes; watch events on controller and clients.
• Verify Proxmox tasks log clean stops; NAS shows clean export unmounts.
• After power return, confirm services come back without fsck or ZFS scrub errors.

Hardening & Monitoring

• Alert on battery tests, low runtime, and communication loss.
• Restrict NUT listener to management VLAN; firewall it off from IoT/Guest.
• Quarterly battery self-test; replace batteries on schedule.