Latest LLM Summary

## Daily summary
• RouterOS CPU/Memory: 1.0% / 15.9% (delta 0.0% vs yesterday, within 7-day avg of 1.43% / 15.3%)
• Proxmox `proxnode` memory: 80.7% (delta 0.0% vs yesterday, exceeds 80% threshold; 7-day avg 76.4%)
• WireGuard reliability: `peer3` & `peer5` failed handshakes repeatedly between 07:26–08:20 EET, giving up after 20 retries
• Docker fleet: 45/45 containers running, 0 restarts, 0 unhealthy
• Docker log errors: 503 lines (delta 0 vs yesterday, 1.28x 7-day avg of 393); top sources: `caddy-dyn` (194), `phase-backend` (137)
• Proxmox storage: max utilization 73.1% on `nvme1tb` (delta 0.0%, all pools <80%)
• RouterOS traffic: RX 1.03 TB / TX 1.27 TB (delta 0.0%, ~1.36x / 1.23x 7-day avg)
• DHCP: 59 active leases, 67 blocked, 0 new in 24h
• PBS backup status: missing from payload; last successful backup cannot be verified

## Actions
- proxmox: investigate high memory on `proxnode`: `ssh proxnode 'free -h && ps aux --sort=-%mem | head -10'`
- routeros: verify WireGuard peer routing/firewall: `ssh routeros '/interface wireguard peers print'`
- docker: review high-error container logs: `docker logs --tail 50 caddy-dyn && docker logs --tail 50 phase-backend`
- pbs: confirm backup job status: `ssh pbs-server 'pbs datastore list'`

## FYI (normal but noteworthy)
• RouterOS interface errors static at 61,983 (no daily delta); monitor for growth
• Firewall dstnat logs show routine SYN scans to ports 80/443 from external IPs; no drop spikes detected
• Proxmox VMs `cachyos`, `lemonade-ubuntu`, `llama-vulkan-bookworm`, `llm-dual-gpu`, `win11-gpu`, `win11-pro` remain stopped as expected