Self-hosted services
Internal tools, bots, and web services on cluster compute I control.
k3s
Traefik
Longhorn
nginx
▲ Self-hosted · k3s · Resilient infrastructure
Skylar Network Systems
Infrastructure
that knows itself.
A private lab running Kubernetes, self-hosted services, telemetry, and automation — built to be observable and honest about its own state.
What is this?
A private cloud, built at home.
A cluster of small computers running self-hosted services, monitoring, and automation — maintained by one person, from home.
Real hardware, real networking
Physical machines in a redundant cluster — self-hosted, not rented.
Monitored end-to-end
Metrics, logs, and service health gathered continuously. Problems surface before they're noticed.
Private and controlled
TLS everywhere, Cloudflare Tunnel, no open ports. Data stays on hardware I own.
Capabilities
Infrastructure with a nervous system.
Connected infrastructure — understood at a glance, repaired with intention.
Monitoring & telemetry
Uptime, power draw, logs, and service health — gathered continuously.
Prometheus
Grafana
Loki
Alertmanager
Security posture
Layered access control, TLS everywhere, firewall rules, and log review. Recovery paths are documented and exercised, not assumed to work.
TLS / mTLS
Cloudflare
fail2ban
audit logs
Backup integrity
Scheduled, verified, and periodically tested restores — not just set and forgotten.
restic
rustic
offsite S3
checksums
Automation
Repeated repairs become repeatable tools — scripts, playbooks, and CronJobs.
Ansible
Python
bash
k8s CronJobs
Operational dashboards
Raw system noise translated into language that helps decide what to do next. Dashboards should reduce anxiety, not manufacture it.
Grafana
skynet CLI
runbooks
Current stack
What's actually running.
Hardware and software as of the last update. The stack evolves.
Hardware
HP EliteDesk 800 G3 Mini — 3 nodes
i5-7500T · 16 GB RAM · 256 GB NVMe · Debian 13 · GeeekPi RackMate T2 12U 10" cabinet
NETGEAR GS308EP switch
8-port PoE+ Gigabit · rack-mounted · direct LAN to all nodes
GL.iNet GL-X3000 Spitz AX
5G/LTE gateway · Wi-Fi 6 · edge router · replaced Protectli V1410
Linux workstations
Daily compute · build hosts · cluster access
Tripp Lite UPS + ElecVoztile rack PDU
BC350R 350VA · 10" 1U PDU · 4 outlets · 1020J surge · 15A overload switch
Ubiquiti UniFi G5 Ultra
4K PoE network camera · managed via UniFi Protect on Cloud Key Gen2+
Ubiquiti UniFi Cloud Key Gen2 Plus
UCK-G2-PLUS · 1 TB Toshiba · UniFi Network + Protect controller
Software platform
k3s · Traefik · Longhorn
Lightweight Kubernetes · ingress routing · replicated block storage
Prometheus · Grafana · Loki
Metrics collection · dashboards · log aggregation · alerting
Cloudflare Tunnel · cert-manager
Zero-trust edge · automatic TLS · no open inbound ports
restic / rustic · offsite S3
Encrypted snapshots · checksum-verified · tested restores
The lab
A household of machines.
Compute, networking, storage, and monitoring as connected systems — understood, not just kept running.
HP EliteDesk cluster
GeeekPi RackMate T2
Protectli V1410 node helper
Linux workstations
UPS telemetry
central logging
TLS edge
Design principles
-
Observe first. Gather evidence before touching the system. Log before you guess.
-
Automate carefully. Turn repeated repairs into repeatable, reviewable tools — not magic scripts.
-
Keep recovery visible. Backups, logs, and service state should be easy to locate and verify, not just scheduled and forgotten.
-
Prefer legibility over cleverness. Systems and scripts should be readable by a future version of me who forgot the context.
Signal Board
System telemetry at a glance.
Tunnel latency, power draw, backup health, and service readiness — live data, updated every 15 minutes.
Tunnel Latency
ms
Cloudflare Tunnel · edge path · 7-sample window
Power Draw
W
7-sample window
Backup Freshness
score / 100
restic snapshots · weekly view · higher is better
Service Health
Core services · from /data/service-health.json
-
TunnelConnected
-
TraefikRouting
-
k3sReady
-
BackupsRecent
Edge Security
24 h
Cloudflare WAF · managed rules · last 24 hours
—
blocked
—
challenged
—
total events
Status
System snapshot.
Live data from the cluster — updated every 15 minutes.
Cluster nodes
Loading…
—
—
Ingress / Traefik
Loading…
—
—
Storage (Longhorn)
Loading…
—
—
Backups
Loading…
—
—
Monitoring
Loading…
—
—
Power / UPS
Loading…
—
—Visitors
Where visitors are coming from.
Top countries by request volume — last 30 days via Cloudflare analytics.
Top countries
Origin map
Connect
Questions, collaboration, or just curious?
Personal project. Questions about homelab infrastructure or anything here — reach out directly.