Skip to main content

Infrastructure Monitoring & Alerting Plan

Executive Summary

This document establishes the architecture, service inventory, and phased rollout plan for comprehensive telemetry, uptime monitoring, and alerting across the self-hosted infrastructure (ul.drheck.dev) and external nodes (rng.drheck.dev).

1. Current Infrastructure Inventory

Primary Host (ul.drheck.dev)

    Hardware: AMD Ryzen Threadripper PRO 5955WX (16 Cores / 32 Threads), 128 GB RAM, 468 GB NVMe (/) with extensive storage pools mounted under /data. Edge Proxy: Nginx reverse proxy with SSL termination and domain routing for *.drheck.dev. Multi-Container Docker Compose Stacks (/data/docker/):
      Identity & Security: Authentik (Server, Worker, LDAP), Step-CA. Media & Gaming: Immich (Server + ML), Funkwhale, RomM (+ MariaDB). Productivity & Docs: Nextcloud + Collabora CODE, BookStack (+ MariaDB), Penpot (Frontend, Backend, Exporter, Postgres, Redis), WIPStudio (+ MinIO). Infrastructure & Web Apps: Webshell, Anchor, YOURLs (+ MariaDB), Valkey, SControl. Rootless Podman Quadlet Services (docker-data@ / /home/docker-data/.config/containers/systemd/):
        Identity & Media: Vaultwarden, Navidrome, Sopro TTS, Demucs Stem Splitter. Collaborative & Standalone Web Apps: AudioMass, BentoPDF, Drawpile, Excalidraw, TLDraw, WebRcade, Wick. Native Systemd Services: Nginx, HAProxy, Unbound, PostgreSQL, Forgejo, Jellyfin, Plex Media Server, Kavita, Minecraft (mc-manager + minecraft@*), Transmission-daemon, CrowdSec, Kanji.

        External VPS (rng.drheck.dev)

          Hardware: Debian 12 (1 GB RAM, 16 GB available on /home storage mount). Role: Off-site canary sentinel, external HTTP/SSL prober, and out-of-band alert dispatcher.

          2. Monitoring Architecture & Topology

                                ┌───────────────────────────────────────────┐
                      │          Off-Site VPS (rng.drheck.dev)    │
                      │  - Uptime Kuma (Blackbox & SSL Prober)    │
                      │  - External WAN/DNS/Port Health Probes    │
                      │  - Independent Alert Routing (ntfy/Pushover)│
                      └─────────────────────┬─────────────────────┘
                                            │ Probes *.drheck.dev (HTTPS/IPv6/IPv4)
                                            ▼
┌────────────────────────────────────────────────────────────────────────────────────────┐
│  Primary Host (ul.drheck.dev)                                                          │
│                                                                                        │
│  [ Telemetry Exporters ]                                                               │
│   ├── node_exporter         (Host CPU, RAM, NVMe I/O, Network, Systemd)                │
│   ├── cAdvisor              (Per-container CPU, Memory, Network, Restart Loops)        │
│   ├── nginx-prometheus      (Request rates, 4xx/5xx error spikes, upstream latencies)  │
│   ├── postgres_exporter     (DB connections, query latency, lock monitoring)           │
│   └── smartctl_exporter     (NVMe wear, drive temperatures, hardware health)           │
│                                                                                        │
│  [ Time-Series Storage & Dashboards ]                                                  │
│   ├── VictoriaMetrics       (High-efficiency TSDB, low memory footprint, PromQL)       │
│   └── Grafana               (Unified dashboards & visual panels)                       │
│                                                                                        │
│  [ Alert Engine ]                                                                      │
│   └── Alertmanager          (Routes internal metrics -> Telegram / ntfy / Matrix)      │
└────────────────────────────────────────────────────────────────────────────────────────┘

          3. Core Component Roles

          A. External Canary (Uptime Kuma on rng)

            Purpose: Detects total site loss, ISP drops, power outages, and certificate expiration before users notice. Why rng: Because rng is outside the local network, it accurately detects external DNS, IPv6 routing, and gateway failures that local monitors cannot detect. Configuration: Probes all public subdomains (wiki.drheck.dev, auth.drheck.dev, immich.drheck.dev, cloud.drheck.dev, etc.) on 60-second intervals.

            B. Host & Container Telemetry (VictoriaMetrics + cAdvisor + Exporters on ul)

              VictoriaMetrics: Selected over standard Prometheus due to its significantly lower memory footprint, superior compression for NVMe endurance, and drop-in PromQL compatibility. cAdvisor: Continuously monitors all 20+ Docker workloads, capturing CPU throttling, memory leaks, and container restart loops. node_exporter: Monitors host-level thread saturation, memory usage, network interfaces (eth77), and disk utilization across mounts.

              C. Unified Visualization (Grafana on ul)

                Preconfigured Dashboards:
                  Host Overview: 32-thread CPU load, RAM allocation, NVMe I/O, network throughput. Docker Fleet: Container status, memory limits, restart counters, per-service resource breakdown. Web Traffic & Edge: Nginx request rates, response latency percentiles (p95/p99), HTTP status distribution (2xx/3xx/4xx/5xx). Database Health: PostgreSQL and MariaDB active connections, buffer hits, query response times.

                  4. Phased Rollout Plan

                  Phase 1: External Sentinel (rng.drheck.dev)

                    Deploy Uptime Kuma under /home/luke/docker/uptime-kuma on rng using Docker or rootless Podman. Configure HTTP(s), TCP, and SSL certificate monitors for all public services on drheck.dev. Configure notification webhooks (e.g. ntfy.sh, Telegram, or Gotify) for instant push alerts to mobile devices.

                    Phase 2: Exporter Deployment (ul.drheck.dev)

                      Deploy node_exporter as a native systemd unit or container (listening on 127.0.0.1:9100). Deploy cAdvisor in Docker (/data/docker/cadvisor/compose.yaml) with read-only mounts to /var/run/docker.sock, /sys, and /var/lib/docker. Add smartctl_exporter for continuous NVMe drive telemetry.

                      Phase 3: Time-Series DB & Grafana Stack (ul.drheck.dev)

                        Create compose stack under /data/docker/monitoring/compose.yaml with:
                          victoriametrics (TSDB storage) grafana (Dashboard UI) vmagent or Prometheus scraping configuration Configure scrape jobs for node_exporter, cAdvisor, Nginx metrics, and database exporters. Expose Grafana behind Nginx at monitor.drheck.dev (secured with Authentik SSO).

                          Phase 4: Alert Rules & Thresholds

                          Define critical alert rules in VictoriaMetrics / Alertmanager:

                            Disk Space: Warning at >80% utilization; Critical at >90% on / or any /data mount. Service Outage: Any service unreachable for > 2 probe cycles (120s). Container Crashes: Container restart count > 3 in a 10-minute window. Host Resource Saturation: RAM available < 10% or continuous high load average. SSL Certificates: Alert if certificate validity < 14 days remaining.

                            Phase 5: Verification & Drills

                              Perform synthetic service termination drill (e.g. test container stop) to verify alert firing and recovery notification. Verify dashboard graphs and metric retention policies.